Instructions to use WaveCut/LingBot-Video-MoE-30B-A3B-SDNQ-uint4-static with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use WaveCut/LingBot-Video-MoE-30B-A3B-SDNQ-uint4-static with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("WaveCut/LingBot-Video-MoE-30B-A3B-SDNQ-uint4-static", dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
Add files using upload-large-folder tool
Browse files- README.md +69 -198
- SHA256SUMS +160 -0
- benchmark/audits/hub-weight-audit.json +178 -0
- benchmark/audits/numerical-smoke.json +23 -0
- benchmark/audits/source-audit.json +18 -0
- benchmark/audits/video-encoding-audit.json +139 -0
- benchmark/audits/video-encoding-repair.json +29 -0
- benchmark/base/original/command.json +21 -0
- benchmark/base/original/metrics.csv +6 -0
- benchmark/base/original/metrics.jsonl +5 -0
- benchmark/base/original/summary.json +370 -0
- benchmark/base/sdnq/metrics.csv +6 -0
- benchmark/base/sdnq/summary.json +370 -0
- benchmark/coverage/refiner.json +0 -0
- benchmark/coverage/transformer.json +0 -0
- benchmark/environment/environment.json +14 -0
- benchmark/environment/nvidia-smi-q.txt +258 -0
- benchmark/environment/pip-freeze.txt +170 -0
- benchmark/original.metrics.csv +6 -0
- benchmark/original.metrics.jsonl +5 -0
- benchmark/paired.metrics.csv +6 -0
- benchmark/paired.metrics.jsonl +5 -0
- benchmark/prompts/01_arctic_research_drone.json +95 -0
- benchmark/prompts/02_macro_pastry_glaze.json +73 -0
- benchmark/prompts/03_rain_market_lanterns.json +99 -0
- benchmark/prompts/04_underwater_manta.json +95 -0
- benchmark/prompts/05_factory_battery_cells.json +111 -0
- benchmark/refiner.original.metrics.csv +2 -0
- benchmark/refiner.original.metrics.jsonl +1 -0
- benchmark/refiner.paired.metrics.csv +2 -0
- benchmark/refiner.paired.metrics.jsonl +1 -0
- benchmark/refiner.sdnq.metrics.csv +2 -0
- benchmark/refiner.sdnq.metrics.jsonl +1 -0
- benchmark/sdnq.metrics.csv +6 -0
- benchmark/sdnq.metrics.jsonl +5 -0
- benchmark/shared/01_arctic_research_drone.meta.json +29 -0
- benchmark/smokes/moe-refiner-sdnq-standard.json +17 -0
- benchmark/smokes/moe-refiner-sdnq-standard.mp4 +0 -0
- benchmark/smokes/moe-sdnq-model.json +17 -0
- benchmark/smokes/moe-sdnq-model.mp4 +0 -0
- benchmark/smokes/moe-sdnq-sequential.json +17 -0
- benchmark/smokes/moe-sdnq-sequential.mp4 +0 -0
- benchmark/smokes/moe-sdnq-standard.json +17 -0
- benchmark/smokes/moe-sdnq-standard.mp4 +0 -0
- benchmark/summary.json +1218 -0
- lingbot_sdnq_runtime/__init__.py +19 -0
- lingbot_sdnq_runtime/__pycache__/__init__.cpython-314.pyc +0 -0
- prompts.json +536 -0
- quantization_manifest.json +0 -0
README.md
CHANGED
|
@@ -1,242 +1,113 @@
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
---
|
| 4 |
|
| 5 |
-
# LingBot
|
| 6 |
|
| 7 |
-
|
| 8 |
|
| 9 |
-
|
| 10 |
|
| 11 |
-
##
|
| 12 |
|
| 13 |
-
|
| 14 |
-
* **📦 Data Engine**: Trained on massive web videos integrated with **70,000+ hours** of embodied data.
|
| 15 |
-
* **⚖️ Multi Reward System**: Rewarded for **high aesthetics**, **physical rationality**, and **task completion**.
|
| 16 |
|
| 17 |
-
|
|
|
|
|
|
|
|
|
|
| 18 |
|
| 19 |
-
-
|
| 20 |
|
| 21 |
-
##
|
| 22 |
|
| 23 |
-
|
| 24 |
-
| --- | --- | --- | --- |
|
| 25 |
-
| ⚡ LingBot-Video-Dense | Dense (1.3B) | T2I, T2V, TI2V | 🤗 [Huggingface](https://huggingface.co/robbyant/lingbot-video-dense-1.3b) 🤖 [ModelScope](https://www.modelscope.cn/models/Robbyant/lingbot-video-dense-1.3b) |
|
| 26 |
-
| 💪 LingBot-Video-MoE | MoE (30B-A3B) + Refiner | T2I, T2V, TI2V, Refinement | 🤗 [Huggingface](https://huggingface.co/robbyant/lingbot-video-moe-30b-a3b) 🤖 [ModelScope](https://www.modelscope.cn/models/Robbyant/lingbot-video-moe-30b-a3b) |
|
| 27 |
-
| 📝 LingBot-Video-Rewriter-Base | Qwen3.6-27B official | Prompt rewriter (Expand) | 🤗 [Huggingface](https://huggingface.co/Qwen/Qwen3.6-27B) 🤖 [ModelScope](https://www.modelscope.cn/models/Qwen/Qwen3.6-27B) |
|
| 28 |
-
| 📝 LingBot-Video-Rewriter-Adapter | Qwen3.6-27B LoRA | Prompt rewriter (Json) | 🤗 [Huggingface](https://huggingface.co/robbyant/lingbot-video-rewriter-lora) 🤖 [ModelScope](https://www.modelscope.cn/models/Robbyant/lingbot-video-rewriter-lora) |
|
| 29 |
|
| 30 |
-
|
|
|
|
|
|
|
|
|
|
| 31 |
|
| 32 |
-
|
| 33 |
|
| 34 |
-
|
| 35 |
|
| 36 |
-
|
| 37 |
-
git clone https://github.com/Robbyant/lingbot-video
|
| 38 |
-
cd lingbot-video
|
| 39 |
-
|
| 40 |
-
python -m venv .venv
|
| 41 |
-
source .venv/bin/activate
|
| 42 |
-
python -m pip install -U pip
|
| 43 |
-
|
| 44 |
-
# Base requirements cover direct DiT inference and rewriter --backend transformers.
|
| 45 |
-
pip install -r requirements.txt
|
| 46 |
-
pip install -e .
|
| 47 |
-
```
|
| 48 |
-
|
| 49 |
-
> **💡 Rewriter deployment**: the bundled rewriter uses the single-process
|
| 50 |
-
> `transformers` backend. For higher throughput, deploy the VLM yourself and call
|
| 51 |
-
> it through an OpenAI-compatible API. Preserve the two-stage semantics: step 1
|
| 52 |
-
> must use the base VLM without the rewriter LoRA, while step 2 must use the same
|
| 53 |
-
> base VLM with the rewriter LoRA enabled. This can be implemented with two
|
| 54 |
-
> endpoints, or with one server that can select the adapter per request. See
|
| 55 |
-
> [vLLM](https://docs.vllm.ai) / [SGLang](https://docs.sglang.ai) official docs.
|
| 56 |
-
|
| 57 |
-
Install the optional SGLang dependencies only when using SGLang Diffusion or the
|
| 58 |
-
fused / FP8 MoE runtime:
|
| 59 |
|
| 60 |
-
```
|
| 61 |
-
python -m pip install --no-deps -r requirements-sglang.txt
|
| 62 |
-
```
|
| 63 |
-
|
| 64 |
-
Recommended runtime versions:
|
| 65 |
|
| 66 |
-
|
| 67 |
-
| --- | --- |
|
| 68 |
-
| `Python` | `>=3.10` |
|
| 69 |
-
| `torch` | `2.12.0.dev20260220+cu130` (recommended) |
|
| 70 |
-
| `torchvision` | `0.26.0.dev20260220+cu130` (recommended) |
|
| 71 |
-
| `transformers` | `5.8.1` |
|
| 72 |
-
| `diffusers` | `0.39.0` |
|
| 73 |
-
| `peft` | `0.19.1` |
|
| 74 |
-
| `json_repair` | `>=0.30` |
|
| 75 |
-
| `decord` | `>=0.6.0` |
|
| 76 |
-
| `safetensors` | `>=0.4.5` |
|
| 77 |
|
| 78 |
-
|
| 79 |
|
| 80 |
-
|
|
|
|
|
|
|
|
|
|
| 81 |
|
| 82 |
-
|
| 83 |
-
not casual natural-language prompts. The recommended public workflow is:
|
| 84 |
|
| 85 |
-
|
| 86 |
-
Prompt Rewriter.
|
| 87 |
-
For TI2V, pass the same first frame to the rewriter.
|
| 88 |
-
2. Run Auto Negative by
|
| 89 |
-
default to prune the negative prompt for this specific caption.
|
| 90 |
-
3. Run the unified inference runner with `--prompt_json` and select direct
|
| 91 |
-
diffusers or SGLang Diffusion through `--backend`.
|
| 92 |
|
| 93 |
-
|
| 94 |
|
| 95 |
-
|
| 96 |
-
- `sglang`: SGLang Diffusion path. If the optional SGLang package is not
|
| 97 |
-
installed, it automatically falls back to direct diffusers and prints a
|
| 98 |
-
warning. Install `requirements-sglang.txt` to enable the SGLang runtime.
|
| 99 |
|
| 100 |
-
|
| 101 |
-
refiner DiT on GPU. This reduces GPU memory pressure after loading, but each
|
| 102 |
-
rank still constructs the transformer on host memory before FSDP sharding; make
|
| 103 |
-
sure the machine has enough system RAM for large MoE checkpoints.
|
| 104 |
-
|
| 105 |
-
```bash
|
| 106 |
-
# Model root (released Dense or MoE package) and rewriter weights.
|
| 107 |
-
export MODEL_DIR="<path_to_lingbot-video-model>"
|
| 108 |
-
export REWRITER_BASE_MODEL="<path_to_rewriter_base_vlm>"
|
| 109 |
-
export REWRITER_ADAPTER="<path_to_rewriter_lora>"
|
| 110 |
-
|
| 111 |
-
python rewriter/inference.py --backend transformers --mode t2v \
|
| 112 |
-
--prompt "<plain_user_prompt>" --duration 5 --output prompt.json
|
| 113 |
-
|
| 114 |
-
# Recommended Auto Negative block. If skipped, remove --negative_prompt_json from
|
| 115 |
-
# the DiT inference command.
|
| 116 |
-
python rewriter/auto_negative.py --backend transformers --mode t2v \
|
| 117 |
-
--caption prompt.json --output negative.json
|
| 118 |
-
|
| 119 |
-
export BACKEND=diffusers # or: sglang
|
| 120 |
-
|
| 121 |
-
python scripts/inference.py \
|
| 122 |
-
--backend "$BACKEND" \
|
| 123 |
-
--model_dir "$MODEL_DIR" \
|
| 124 |
-
--run_refiner \
|
| 125 |
-
--mode t2v \
|
| 126 |
-
--prompt_json prompt.json \
|
| 127 |
-
--negative_prompt_json negative.json \
|
| 128 |
-
--output "<output_dir>/base.mp4" \
|
| 129 |
-
--refiner_output "<output_dir>/refined.mp4" \
|
| 130 |
-
--height 480 \
|
| 131 |
-
--width 832 \
|
| 132 |
-
--fps 24 \
|
| 133 |
-
--steps 40 \
|
| 134 |
-
--refiner_steps 8 \
|
| 135 |
-
--guidance_scale 3 \
|
| 136 |
-
--refiner_guidance_scale 3 \
|
| 137 |
-
--shift 3 \
|
| 138 |
-
--refiner_shift 3 \
|
| 139 |
-
--transformer_dtype bf16 \
|
| 140 |
-
--text_encoder_dtype bf16 \
|
| 141 |
-
--vae_dtype fp32 \
|
| 142 |
-
--refiner_vae_dtype fp32 \
|
| 143 |
-
--reuse_condition_features
|
| 144 |
-
```
|
| 145 |
|
| 146 |
-
|
| 147 |
-
your environment and model path first:
|
| 148 |
|
| 149 |
```bash
|
| 150 |
-
|
| 151 |
-
|
| 152 |
-
|
| 153 |
-
export MOE_MODEL_DIR="<path_to_lingbot-video-moe>"
|
| 154 |
```
|
| 155 |
|
| 156 |
-
|
| 157 |
-
base generation only.
|
| 158 |
|
| 159 |
```bash
|
| 160 |
-
|
| 161 |
-
MODEL_DIR="$DENSE_MODEL_DIR" ./scripts/single-gpu/run_dense_t2v.sh
|
| 162 |
-
MODEL_DIR="$DENSE_MODEL_DIR" ./scripts/single-gpu/run_dense_ti2v.sh
|
| 163 |
-
|
| 164 |
-
MODEL_DIR="$MOE_MODEL_DIR" ./scripts/single-gpu/run_moe_t2i.sh
|
| 165 |
-
MODEL_DIR="$MOE_MODEL_DIR" ./scripts/single-gpu/run_moe_t2v.sh
|
| 166 |
-
MODEL_DIR="$MOE_MODEL_DIR" ./scripts/single-gpu/run_moe_ti2v.sh
|
| 167 |
```
|
| 168 |
|
| 169 |
-
|
| 170 |
-
scripts, plus CP8 and FSDP:
|
| 171 |
-
|
| 172 |
-
```bash
|
| 173 |
-
MODEL_DIR="$DENSE_MODEL_DIR" ./scripts/multi-gpus-no-refiner/run_dense_t2i_fsdp_cp8.sh
|
| 174 |
-
MODEL_DIR="$DENSE_MODEL_DIR" ./scripts/multi-gpus-no-refiner/run_dense_t2v_fsdp_cp8.sh
|
| 175 |
-
MODEL_DIR="$DENSE_MODEL_DIR" ./scripts/multi-gpus-no-refiner/run_dense_ti2v_fsdp_cp8.sh
|
| 176 |
|
| 177 |
-
|
| 178 |
-
|
| 179 |
-
|
| 180 |
-
```
|
| 181 |
|
| 182 |
-
|
| 183 |
-
|
| 184 |
-
exercise SGLang Diffusion. MoE multi-GPU T2V/TI2V scripts additionally run the
|
| 185 |
-
refiner.
|
| 186 |
|
| 187 |
-
|
| 188 |
-
MODEL_DIR="$DENSE_MODEL_DIR" ./scripts/multi-gpus/run_dense_t2i_fsdp_cp8.sh
|
| 189 |
-
MODEL_DIR="$DENSE_MODEL_DIR" ./scripts/multi-gpus/run_dense_t2v_fsdp_cp8.sh
|
| 190 |
-
MODEL_DIR="$DENSE_MODEL_DIR" ./scripts/multi-gpus/run_dense_ti2v_fsdp_cp8.sh
|
| 191 |
|
| 192 |
-
|
| 193 |
-
|
| 194 |
-
MODEL_DIR="$MOE_MODEL_DIR" ./scripts/multi-gpus/run_moe_ti2v_refiner_fsdp_cp8.sh
|
| 195 |
```
|
| 196 |
|
| 197 |
-
|
| 198 |
-
`IMAGE`, `OUT_DIR`, `HEIGHT`, `WIDTH`, `STEPS`, `GUIDANCE_SCALE`, `SHIFT`,
|
| 199 |
-
`SEED`, `FPS`, `BACKEND`, and `PYTHON_BIN`. Refiner scripts also accept
|
| 200 |
-
`REFINER_HEIGHT`, `REFINER_WIDTH`, `REFINER_STEPS`,
|
| 201 |
-
`REFINER_GUIDANCE_SCALE`, `REFINER_SHIFT`, `REFINER_T_THRESH`, and
|
| 202 |
-
`REFINER_SIGMA_TAIL_STEPS`. MoE scripts default to grouped expert execution
|
| 203 |
-
(`LINGBOT_MOE_EXPERT_BACKEND=grouped_mm`).
|
| 204 |
-
|
| 205 |
-
See English Docs or 中文文档 for the
|
| 206 |
-
detailed prompt rewrite, auto-negative, TI2V, base-only/refiner, distributed
|
| 207 |
-
SGLang, and speed-first FP8 workflows.
|
| 208 |
|
| 209 |
-
##
|
| 210 |
|
| 211 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 212 |
|
| 213 |
-
|
| 214 |
|
| 215 |
-
|
| 216 |
-
|
| 217 |
-
|
| 218 |
-
|
| 219 |
-
|
| 220 |
-
|
| 221 |
-
| HunyuanVideo 1.5 | ✅ | 0.460 | 0.442 | 0.316 | 0.312 | 0.438 | 0.364 | 0.513 | 0.526 | 0.634 | 0.595 |
|
| 222 |
-
| Wan 2.6 | ❌ | <u>0.607</u> | 0.546 | **0.656** | <u>0.479</u> | 0.514 | **0.531** | **0.666** | **0.681** | 0.723 | 0.667 |
|
| 223 |
-
| Seedance 1.5 pro | ❌ | 0.584 | <u>0.577</u> | 0.495 | **0.484** | 0.570 | 0.470 | <u>0.648</u> | <u>0.641</u> | 0.680 | **0.692** |
|
| 224 |
-
| Veo 3 | ❌ | 0.563 | 0.521 | 0.508 | 0.430 | 0.530 | 0.504 | 0.634 | 0.610 | 0.689 | 0.637 |
|
| 225 |
-
|
| 226 |
-
|
| 227 |
-
*Note: **Bold** indicates the best performance, and <u>underline</u> indicates the second best.*
|
| 228 |
-
|
| 229 |
-
## ⚖️ License
|
| 230 |
-
This project is licensed under the Apache 2.0 License. Please refer to the [LICENSE file](LICENSE) for the full text, including details on rights and restrictions.
|
| 231 |
-
|
| 232 |
-
## 📚 Citation
|
| 233 |
-
If you find this work useful for your research, please cite our paper:
|
| 234 |
-
|
| 235 |
-
```bibtex
|
| 236 |
-
@article{lingbot-video,
|
| 237 |
-
title = {Scaling Mixture-of-Experts Video Pretraining for Embodied Intelligence},
|
| 238 |
-
author = {Shuailei Ma and Jiaqi Liao and Xinyang Wang and Jingjing Wang and Chaoran Feng and Zijing Hu and Chong Bao and Zichen Xi and Yuqi Gan and Weisen Wang and Yanhong Zeng and Qin Zhao and Zifan Shi and Wei Wu and Hao Ouyang and Qiuyu Wang and Shangzhan Zhang and Jiahao Shao and Yipengjing Sun and Liangxiao Hu and Lunke Pan and Nan Xue and Kecheng Zheng and Yinghao Xu and Xing Zhu and Yujun Shen and Ka Leong Cheng},
|
| 239 |
-
journal={arXiv preprint arXiv:2607.xxxxx},
|
| 240 |
-
year = {2026}
|
| 241 |
-
}
|
| 242 |
-
```
|
|
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
+
library_name: diffusers
|
| 4 |
+
pipeline_tag: text-to-video
|
| 5 |
+
base_model: robbyant/lingbot-video-moe-30b-a3b
|
| 6 |
+
tags:
|
| 7 |
+
- lingbot-video
|
| 8 |
+
- sdnq
|
| 9 |
+
- uint4
|
| 10 |
+
- text-to-video
|
| 11 |
---
|
| 12 |
|
| 13 |
+
# LingBot Video MoE 30B-A3B SDNQ UINT4 Static
|
| 14 |
|
| 15 |
+
This is a complete, loadable derivative of [robbyant/lingbot-video-moe-30b-a3b](https://huggingface.co/robbyant/lingbot-video-moe-30b-a3b) with the diffusion transformer stored using static SDNQ UINT4 weights. It is tied to source model revision `f2e538f64afe00cc4ae674db2aeb52e2945edfd5`, LingBot Video code `a2bb04b78edd848500dc27a26e035a95442ae186`, and SDNQ `d841c383ff7be38728d4df829e17af4f15d4fd66` (`v0.2.1-17-gd841c38`).
|
| 16 |
|
| 17 |
+
Text encoder, tokenizer/processor, scheduler, and VAE remain at their upstream precision. Convolutions and embeddings are not quantized. The recipe is `uint4-static-transformer-only-with-3d-expert-adapter`: `weights_dtype=uint4`, auto group size (`group_size=0`), no dynamic quantization, SVD, Hadamard transform, convolution quantization, or embedding quantization.
|
| 18 |
|
| 19 |
+
## Coverage
|
| 20 |
|
| 21 |
+
Coverage is calculated from the original parameter inventory, not from model-file sizes.
|
|
|
|
|
|
|
| 22 |
|
| 23 |
+
| Component | Total logical params | Quantized params | Original parameter bytes covered | Packed storage | Packed grouped experts |
|
| 24 |
+
| --- | ---: | ---: | ---: | ---: | --- |
|
| 25 |
+
| `transformer` | 30,084,506,176 | 99.8268% | 99.6550% | 17.42 GiB | 28,991,029,248 params / 16.88 GiB stored |
|
| 26 |
+
| `refiner` | 30,084,506,176 | 99.8268% | 99.6550% | 17.42 GiB | 28,991,029,248 params / 16.88 GiB stored |
|
| 27 |
|
| 28 |
+
Both the base transformer's and refiner's raw 3-D `w1`/`w2`/`w3` expert tensors are packed. They are not excluded from the reported coverage. Exact module-level coverage and unquantized tensors are in [`quantization_manifest.json`](quantization_manifest.json) and [`benchmark/coverage`](benchmark/coverage/).
|
| 29 |
|
| 30 |
+
## Reproducible base benchmark
|
| 31 |
|
| 32 |
+
All five pairs use identical prompts, negative prompt, seeds 4201-4205, scheduler inputs, 832x480 dimensions, 73 frames, 24 fps, 40 steps, guidance 3.0, shift 3.0, `batch_cfg=False`, and `null_cond_clone_zero=False`. Resources were sampled every 250 ms from `/proc`, `psutil`, and `nvidia-smi`.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 33 |
|
| 34 |
+
| Variant | Load (s) | Cold generation (s) | Hot mean (s) | Peak VRAM (MiB) | Peak Torch allocated (MiB) | Process RSS (GiB) | System RAM used (GiB) |
|
| 35 |
+
| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
|
| 36 |
+
| Original BF16 | 50.29 | 140.90 | 139.63 | 126274 | 112470 | 3.00 | 86.52 |
|
| 37 |
+
| SDNQ UINT4 | 43.81 | 159.77 | 158.62 | 42504 | 36068 | 3.08 | 129.39 |
|
| 38 |
|
| 39 |
+
Observed base peak-VRAM reduction: **66.34%**. Timing and memory are measurements on the environment recorded in [`benchmark/environment`](benchmark/environment/), not universal performance claims.
|
| 40 |
|
| 41 |
+
Frame-aligned aggregate quality across the five pairs: MAE `0.160701`, RMSE `0.227453`, PSNR `13.561 dB`, SSIM `0.573403`, LPIPS-Alex `0.484469`.
|
| 42 |
|
| 43 |
+

|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 44 |
|
| 45 |
+
The complete contact sheets and side-by-side MP4s are under [`assets/comparison/base`](assets/comparison/base/). Quantized sample MP4s are under [`samples/base`](samples/base/). Raw per-prompt CSV/JSONL, resource samples, commands, ffprobe records, output sizes, and SHA-256 values are under [`benchmark`](benchmark/).
|
|
|
|
|
|
|
|
|
|
|
|
|
| 46 |
|
| 47 |
+
## Refiner A/B
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 48 |
|
| 49 |
+
The refiner pair used the same `502b10f841d96aa101e69421b20083aeb60427c054b42fb0db29c5a6e70824cf` initial latent generated from the same base MP4. Both outputs are 1920x1088, 73 frames at 24 fps, with 8 refiner steps and upstream-default `null_cond_clone_zero=True`.
|
| 50 |
|
| 51 |
+
| Variant | Load (s) | Generation (s) | Peak VRAM (MiB) | Process RSS (GiB) | System RAM used (GiB) |
|
| 52 |
+
| --- | ---: | ---: | ---: | ---: | ---: |
|
| 53 |
+
| Original BF16 | 139.01 | 478.60 | 179178 | 4.26 | 107.63 |
|
| 54 |
+
| SDNQ UINT4 | 105.56 | 460.47 | 98232 | 3.85 | 105.73 |
|
| 55 |
|
| 56 |
+
Frame-aligned refiner quality: MAE `0.014650`, RMSE `0.034053`, PSNR `29.357 dB`, SSIM `0.954816`, LPIPS-Alex `0.063118`.
|
|
|
|
| 57 |
|
| 58 |
+
Visual inspection found the same strong pink/red color clipping and cyan/magenta speckling in both the original-BF16 and SDNQ refiner outputs. Their close frame metrics therefore demonstrate pairwise similarity, not natural-color reconstruction quality; treat this as a shared refiner quality failure in this recorded sample.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 59 |
|
| 60 |
+

|
| 61 |
|
| 62 |
+
Load the refiner by passing `transformer_subfolder="refiner"` to `load_pipeline`.
|
|
|
|
|
|
|
|
|
|
| 63 |
|
| 64 |
+
## Installation and load
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 65 |
|
| 66 |
+
Use the exact pinned dependencies shipped with the repository:
|
|
|
|
| 67 |
|
| 68 |
```bash
|
| 69 |
+
git clone https://huggingface.co/WaveCut/LingBot-Video-MoE-30B-A3B-SDNQ-uint4-static
|
| 70 |
+
cd LingBot-Video-MoE-30B-A3B-SDNQ-uint4-static
|
| 71 |
+
python -m pip install -r runtime-requirements.txt
|
|
|
|
| 72 |
```
|
| 73 |
|
| 74 |
+
The tested MoE backend also needs the pinned SGLang userspace packages without replacing Torch:
|
|
|
|
| 75 |
|
| 76 |
```bash
|
| 77 |
+
python -m pip install --no-deps sglang==0.5.13.post1 apache-tvm-ffi==0.1.9 tilelang==0.1.8 triton==3.6.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 78 |
```
|
| 79 |
|
| 80 |
+
The repository includes the runtime adapter; no unmerged LingBot branch or local hidden file is needed:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 81 |
|
| 82 |
+
```python
|
| 83 |
+
import sys
|
| 84 |
+
from huggingface_hub import snapshot_download
|
|
|
|
| 85 |
|
| 86 |
+
root = snapshot_download("WaveCut/LingBot-Video-MoE-30B-A3B-SDNQ-uint4-static")
|
| 87 |
+
sys.path.insert(0, root)
|
|
|
|
|
|
|
| 88 |
|
| 89 |
+
from lingbot_sdnq_runtime import load_pipeline
|
|
|
|
|
|
|
|
|
|
| 90 |
|
| 91 |
+
pipe = load_pipeline(root, device="cuda")
|
| 92 |
+
# For the MoE refiner: load_pipeline(root, transformer_subfolder="refiner", device="cuda")
|
|
|
|
| 93 |
```
|
| 94 |
|
| 95 |
+
See [`prompts.json`](prompts.json) for the exact A/B inputs and [`benchmark/summary.json`](benchmark/summary.json) for portable metrics. Recorded consumer/offload smoke artifacts: `benchmark/smokes/moe-refiner-sdnq-standard.json`, `benchmark/smokes/moe-refiner-sdnq-standard.mp4`, `benchmark/smokes/moe-sdnq-model.json`, `benchmark/smokes/moe-sdnq-model.mp4`, `benchmark/smokes/moe-sdnq-sequential.json`, `benchmark/smokes/moe-sdnq-sequential.mp4`, `benchmark/smokes/moe-sdnq-standard.json`, `benchmark/smokes/moe-sdnq-standard.mp4`.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 96 |
|
| 97 |
+
## Runtime behavior and limitations
|
| 98 |
|
| 99 |
+
- Generic SDNQ Linear layers use eager BF16 dequantization followed by `F.linear` in the tested Torch 2.8/CUDA 12.8 environment because the current SDNQ Triton quantized-matmul path is incompatible there.
|
| 100 |
+
- Packed MoE experts are dequantized for each expert call and executed by the pinned SGLang Triton fused-MoE path. The adapter does not keep a persistent BF16 expert-weight cache.
|
| 101 |
+
- Static UINT4 materially changes generated pixels. Inspect the published matrices and per-prompt metrics before choosing this derivative for quality-sensitive work.
|
| 102 |
+
- The MoE SDNQ factory prompt is a severe framing/adherence regression: the valid 832x480 MP4 contains a smaller portrait-like factory view centered on a white canvas (MAE `0.338658`). Treat that sample as a quality failure, not as a successful match to the original.
|
| 103 |
+
- Peak residency and speed depend strongly on resolution, frame count, attention backend, offload mode, and GPU. The numbers above describe the exact recorded B200 run only.
|
| 104 |
+
- The Apache-2.0 upstream license is retained. Users remain responsible for evaluating generated content for their application.
|
| 105 |
|
| 106 |
+
## Evidence map
|
| 107 |
|
| 108 |
+
- [`quantization_manifest.json`](quantization_manifest.json): recipe, revisions, per-component and expert coverage.
|
| 109 |
+
- [`prompts.json`](prompts.json): exact structured prompts, negative prompt, seeds, and generation settings.
|
| 110 |
+
- [`benchmark/summary.json`](benchmark/summary.json): portable aggregate benchmark record.
|
| 111 |
+
- [`benchmark/base`](benchmark/base/): unmodified original and SDNQ raw metrics and resource samples.
|
| 112 |
+
- [`benchmark/comparison`](benchmark/comparison/): frame-aligned MAE/RMSE/PSNR/SSIM/LPIPS records.
|
| 113 |
+
- [`SHA256SUMS`](SHA256SUMS): hashes for all published files, including model shards.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
SHA256SUMS
ADDED
|
@@ -0,0 +1,160 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
11ad7efa24975ee4b0c3c3a38ed18737f0658a5f75a0a96787b576a78a023361 .gitattributes
|
| 2 |
+
844e44993d592f6699883d90b6dc26b8e5d95dc56f9e875c0c396a39a40d969a README.md
|
| 3 |
+
42edf48ea051286e77a7f1afbb733b168246af864566b8872ab653f2e93f96ca RUNTIME_ADAPTER.md
|
| 4 |
+
b65a6dfe6efc5a171bbbe67a0a55d4776a15da04d8b846149b0549b894a8c38d assets/comparison/base/contact_sheets/01_arctic_research_drone.webp
|
| 5 |
+
68f35ea8d2234b7ff34a78e3ccceeb283b1ad26461881d70b9065bef82ca6cb8 assets/comparison/base/contact_sheets/02_macro_pastry_glaze.webp
|
| 6 |
+
657f49492f70ab256ffee41ca33a618a73977050479a8d301c804a61ad9e91c6 assets/comparison/base/contact_sheets/03_rain_market_lanterns.webp
|
| 7 |
+
25cafa5c035c77ef7c10a8ed34aa4d918a9ff682f9cf387ed04f28a6be2eb846 assets/comparison/base/contact_sheets/04_underwater_manta.webp
|
| 8 |
+
c8280141ff87012646d681f592df0050325f0ed498644fb9400f021c55f69478 assets/comparison/base/contact_sheets/05_factory_battery_cells.webp
|
| 9 |
+
63d3d461bb087ad1f1f9d41a9a3d51ea8e56457501b26eb01f8a86afefee0c53 assets/comparison/base/original_vs_sdnq_overview.webp
|
| 10 |
+
41841f1a7b8582718df81e184c2ab6da575a3ab6ddf48954bd44e82356450165 assets/comparison/base/paired_videos/01_arctic_research_drone.mp4
|
| 11 |
+
85dc9736202f82cd5cdfe8943d30cfb66b87fcc25ee6a6f14172efa8c1865357 assets/comparison/base/paired_videos/02_macro_pastry_glaze.mp4
|
| 12 |
+
21bcca9a790c32dddaabaefca140bff2ebd41c52b82abbf1330fc2e94c29d674 assets/comparison/base/paired_videos/03_rain_market_lanterns.mp4
|
| 13 |
+
6abef4e55c08d9c90f5d4bf3cfa775c2f72a520e9277e7253d0b0ee056e8a8ba assets/comparison/base/paired_videos/04_underwater_manta.mp4
|
| 14 |
+
3ed371dd50f914e6767130f3963d45becfe4b0fe25a7a2ca7e138bafd78b52b3 assets/comparison/base/paired_videos/05_factory_battery_cells.mp4
|
| 15 |
+
2d1353ecdbcfb514a07d658b7576e0da73894de1f366f6928c7c005193686979 assets/comparison/refiner/contact_sheets/01_arctic_research_drone.webp
|
| 16 |
+
1c60903f42c86777da5149576054bd4c4798e6253acba8794bdf84f756ddd44e assets/comparison/refiner/original_vs_sdnq_overview.webp
|
| 17 |
+
603c8ca01e33b3cb2649b73efa97d6d116d0d5791ee8327d9c87e13d578d5d71 assets/comparison/refiner/paired_videos/01_arctic_research_drone.mp4
|
| 18 |
+
7b7f06d162dcc66f2cb7cd6be53ca542641b61cbb1f2a09a78032487d06b7316 benchmark/audits/hub-weight-audit.json
|
| 19 |
+
4deed17bc182de4fba52bc7038abce2d6feef685b8d738d1e7e471860dbaa699 benchmark/audits/numerical-smoke.json
|
| 20 |
+
e12d195e69f97b498b7012ebf749d1bf6dc641e61fa12725ff4b8cf55b8b4054 benchmark/audits/source-audit.json
|
| 21 |
+
1f545776eb73accbe06b50f82d3596a61532053d4e64fd57efeb7538d0bbcb22 benchmark/audits/video-encoding-audit.json
|
| 22 |
+
f80ef1461f93d1ea1d7ad4761773bb3a043643c6b24110a02656da0dd396afa0 benchmark/audits/video-encoding-repair.json
|
| 23 |
+
6bb1935f95d368300f85a799c7052e08fc979339529f0ed98f2d7764aa33fd9f benchmark/base/original/command.json
|
| 24 |
+
002014b2ba44d86d07f6fd86aa8667a48d895c0ab81748d540187b55ada012ad benchmark/base/original/metrics.csv
|
| 25 |
+
f9f472639536b51bffc7d730211b959a18e23862f2276f2003ea1f3455516f46 benchmark/base/original/metrics.jsonl
|
| 26 |
+
e89590519a625d8455b1174bc7ed5d40dee249a5f753fcb32c5d00007ad6ebae benchmark/base/original/resource_samples/01_arctic_research_drone.jsonl
|
| 27 |
+
7fac43ffffacaa696c69cf42392cece89ad7ab3197223b8601aa1a2f8fd79f02 benchmark/base/original/resource_samples/02_macro_pastry_glaze.jsonl
|
| 28 |
+
93f1f3bde681d3e7e8d85481fe38fcecacd51fbaa70ef78cb8b27c927bb2840d benchmark/base/original/resource_samples/03_rain_market_lanterns.jsonl
|
| 29 |
+
e236d0b7c0319b0d83e78f74ba923e09a364c7a78fcb42858352516acdef105a benchmark/base/original/resource_samples/04_underwater_manta.jsonl
|
| 30 |
+
773032020bb9a304c1a8f0bb7d6b42eff193b31a1d5425790ea0bed09b94f3ac benchmark/base/original/resource_samples/05_factory_battery_cells.jsonl
|
| 31 |
+
b3c83c5e4762ccea9070930afeeed55c6bccbe13ed21d4b1bacb59df75ebaa81 benchmark/base/original/resource_samples/load.jsonl
|
| 32 |
+
bff343ad90b0e4c3aff0d351ae958352b74f74ea294eef27acc648342ae7a2cc benchmark/base/original/summary.json
|
| 33 |
+
6ff2a4f9f133abfed4e25ea8595b367e71000a21956ec298f65972113cb3e35a benchmark/base/sdnq/command.json
|
| 34 |
+
7037f793dbbc9fd295041fdeb6be59406c458cca839241c1636ea585a4160ccb benchmark/base/sdnq/metrics.csv
|
| 35 |
+
a9b9fd6441fd488c9ac5d28269edaaa3f6da06b42b640fc525d942bf2d630fa7 benchmark/base/sdnq/metrics.jsonl
|
| 36 |
+
1aecc6c24a6acfafccca4a33827118bfb328ff5de731f719b7b9b47545f7bd9f benchmark/base/sdnq/resource_samples/01_arctic_research_drone.jsonl
|
| 37 |
+
d82be43050460cd7ca00b09cc2060deb8c16aa7cbbbc7b93014aa4914ceda9fa benchmark/base/sdnq/resource_samples/02_macro_pastry_glaze.jsonl
|
| 38 |
+
898e9adb63cbba76b0f4bad0dac7a84d92cff1443f44084485a1b6ff30495fce benchmark/base/sdnq/resource_samples/03_rain_market_lanterns.jsonl
|
| 39 |
+
913861053b38afa83301f9be292b43d1c8fc731b1d2d2d6a02ba9703819dbf03 benchmark/base/sdnq/resource_samples/04_underwater_manta.jsonl
|
| 40 |
+
fcdd93cd565f364214a9ef0012355992c5992161f073eb1855b80904bdc7c035 benchmark/base/sdnq/resource_samples/05_factory_battery_cells.jsonl
|
| 41 |
+
cbde44804eede2186f9d7753e6135cbaff3df6cf0670af1d92823fa184375fd3 benchmark/base/sdnq/resource_samples/load.jsonl
|
| 42 |
+
91cd9048e873de2e86905b75504cbe328a76365444d841d3994d3f051b0ef027 benchmark/base/sdnq/summary.json
|
| 43 |
+
a5410b14a6d31e3f01b2e70ef6fa1528ab3d5116af402183a06fb334ae7387a7 benchmark/comparison/base/metrics.csv
|
| 44 |
+
909fafe464229dd227eb5cbb969d54056adaf9291c6c623481761f0840123af5 benchmark/comparison/base/metrics.jsonl
|
| 45 |
+
6b79fe4cd4320e00fa151ca054a5d15b3862dc0ae37c6c93357d926992cec708 benchmark/comparison/base/summary.json
|
| 46 |
+
a08efd236f811af4f11e3ffcd68eb0e558d5fc8642b78c1941e3c1cb3374738d benchmark/comparison/refiner/metrics.csv
|
| 47 |
+
c10aae628f48576c5bae3bf15c92849857ef757bf9d5a7be18f74019db832752 benchmark/comparison/refiner/metrics.jsonl
|
| 48 |
+
c347fc98a0268c4242b9832614b591f34db4cae8d6d70877616044dfe62e8b82 benchmark/comparison/refiner/summary.json
|
| 49 |
+
b21bc0b086b03cc9a0000cf7256a712dab2695ffcda7026b076c3abb99dd1eb3 benchmark/coverage/refiner.json
|
| 50 |
+
8bc6949365153e7e5779142a96de01ab23b8979f6c92a4d8ede7eea99e6ae6a5 benchmark/coverage/transformer.json
|
| 51 |
+
d9f5f653536528100b1da486eb2579f5fd2015d92c2043d1f460bb385f04c5a8 benchmark/environment/environment.json
|
| 52 |
+
4e5b86fd706e02f5d8fa4142a1224c4aa21e4626132e0324c2ad593ef3553704 benchmark/environment/nvidia-smi-q.txt
|
| 53 |
+
785818ed73ad5a678b675216a7d33696003884e3d1df9276cd9b1705c5e2a752 benchmark/environment/pip-freeze.txt
|
| 54 |
+
002014b2ba44d86d07f6fd86aa8667a48d895c0ab81748d540187b55ada012ad benchmark/original.metrics.csv
|
| 55 |
+
f9f472639536b51bffc7d730211b959a18e23862f2276f2003ea1f3455516f46 benchmark/original.metrics.jsonl
|
| 56 |
+
97623c36c81076a0de18ad34dcc3a78fe01a3b402db43b1b665577248f965f94 benchmark/paired.metrics.csv
|
| 57 |
+
57b115683b347e565858a6ebf5d24446203f9f674acfb474c2cd9ee0bab40fd9 benchmark/paired.metrics.jsonl
|
| 58 |
+
dc9e321c1128ea7a5925a00299683bbeae7ad60ebdbe54f479679d01944b7e45 benchmark/prompts/01_arctic_research_drone.json
|
| 59 |
+
d05e937b7a82c55195c6ca7deb1ecd5471b96ad752a4ee1c233eac00862538a4 benchmark/prompts/02_macro_pastry_glaze.json
|
| 60 |
+
704450c5ff5a2e509fb306ecaf56ba3d7cb482185437dd80be27ee582d7111eb benchmark/prompts/03_rain_market_lanterns.json
|
| 61 |
+
173c81b1138a563ec56dc6a1f6fd366772901a30164e7748906ef746c319cba9 benchmark/prompts/04_underwater_manta.json
|
| 62 |
+
11f6a49066400a581d9324ffd5318fd68cc21f835ff349a98e01889eb43627e9 benchmark/prompts/05_factory_battery_cells.json
|
| 63 |
+
a203e074551153b5a8e95e01e67643a166c1d2fca8f59abf265b57196dad37b5 benchmark/refiner/original/command.json
|
| 64 |
+
3c4286d5cc9de63cf3627a19de072203f763ac357b6ce8a732b5f87eaf4553ce benchmark/refiner/original/metrics.csv
|
| 65 |
+
007772f6ce92b194b8992c26f376a22317dd89bccf51d00644ae2682f830eac4 benchmark/refiner/original/metrics.jsonl
|
| 66 |
+
fbdd745d6ea7314607ea5e75ed21a847f6768e2293159f76b11b5e1d480170ee benchmark/refiner/original/resource_samples/01_arctic_research_drone.jsonl
|
| 67 |
+
f96e0b8c2d3b86496b01c3d4baa4439c3ff342ef8cff3d7b66b49572bb884e91 benchmark/refiner/original/resource_samples/load.jsonl
|
| 68 |
+
7b6cc4fff7ffb2516775ff6f498f55539153e56ed4f961f64208ddb6b4070b8f benchmark/refiner/original/summary.json
|
| 69 |
+
2e8f5c44c481c9118fc1f644bae60689a6137681b1a219d71967360182f5acc4 benchmark/refiner/sdnq/command.json
|
| 70 |
+
2c5218cad3990812305fd9eeb70d5d8e6a301f655f653ae63c95655b2b125a97 benchmark/refiner/sdnq/metrics.csv
|
| 71 |
+
fc4568de32804bd5c633ce76fcc8e9795195a90e24401e80c44760e8bd22af9b benchmark/refiner/sdnq/metrics.jsonl
|
| 72 |
+
06f99a7846ebaa515288c9e5496778f6f66144353c7ce2b5ab7d24ec780a47d5 benchmark/refiner/sdnq/resource_samples/01_arctic_research_drone.jsonl
|
| 73 |
+
28b2727cb9f69a30f4a7e597bb6b471507e6cb92f629b1df20e75ea7ba919657 benchmark/refiner/sdnq/resource_samples/load.jsonl
|
| 74 |
+
4ef6c64fc83ddaebcc8a5db1ec1d749ef6367775099c0c273b870a29c7050caf benchmark/refiner/sdnq/summary.json
|
| 75 |
+
3c4286d5cc9de63cf3627a19de072203f763ac357b6ce8a732b5f87eaf4553ce benchmark/refiner.original.metrics.csv
|
| 76 |
+
007772f6ce92b194b8992c26f376a22317dd89bccf51d00644ae2682f830eac4 benchmark/refiner.original.metrics.jsonl
|
| 77 |
+
02f71bc3609694cfdfb0f5a24c86881e3bc6824c36f7964054f94d7b143e2084 benchmark/refiner.paired.metrics.csv
|
| 78 |
+
c52f941a1759332a5eb459edbae8001ebb133a420de8a203c0b8eae6f5ad8827 benchmark/refiner.paired.metrics.jsonl
|
| 79 |
+
2c5218cad3990812305fd9eeb70d5d8e6a301f655f653ae63c95655b2b125a97 benchmark/refiner.sdnq.metrics.csv
|
| 80 |
+
fc4568de32804bd5c633ce76fcc8e9795195a90e24401e80c44760e8bd22af9b benchmark/refiner.sdnq.metrics.jsonl
|
| 81 |
+
7037f793dbbc9fd295041fdeb6be59406c458cca839241c1636ea585a4160ccb benchmark/sdnq.metrics.csv
|
| 82 |
+
a9b9fd6441fd488c9ac5d28269edaaa3f6da06b42b640fc525d942bf2d630fa7 benchmark/sdnq.metrics.jsonl
|
| 83 |
+
b6a305a26f8b623c05628728daf019bc104938edc0d6eb26ffa929c6c1222ed6 benchmark/shared/01_arctic_research_drone.meta.json
|
| 84 |
+
502b10f841d96aa101e69421b20083aeb60427c054b42fb0db29c5a6e70824cf benchmark/shared/01_arctic_research_drone.safetensors
|
| 85 |
+
8edafa5c633abd72e66da6c52bd1bd7ca64b919106c5cb15ac4cf13be6db4ca5 benchmark/smokes/moe-refiner-sdnq-standard.json
|
| 86 |
+
4f622465321020f9d63f7efc99416b4f7b8d8e3e48cd7cf2691db904ea052101 benchmark/smokes/moe-refiner-sdnq-standard.mp4
|
| 87 |
+
dce320a801062238301a06f88bb7c0dd7cbca3289bd991025102f362f1bb382d benchmark/smokes/moe-sdnq-model.json
|
| 88 |
+
6765bd8eb52a955c796b5320c29461fe1f975ecc9f1f4ba7d65e5e5517768c7f benchmark/smokes/moe-sdnq-model.mp4
|
| 89 |
+
a0ec161df814995cff93019c1371c21c9e457b7cde7b3149ea445e3d2b913137 benchmark/smokes/moe-sdnq-sequential.json
|
| 90 |
+
6765bd8eb52a955c796b5320c29461fe1f975ecc9f1f4ba7d65e5e5517768c7f benchmark/smokes/moe-sdnq-sequential.mp4
|
| 91 |
+
a23c8676453f4dfb4a1473fbbac6c3a48fcc2ae58932168a341d7037706d64f2 benchmark/smokes/moe-sdnq-standard.json
|
| 92 |
+
6765bd8eb52a955c796b5320c29461fe1f975ecc9f1f4ba7d65e5e5517768c7f benchmark/smokes/moe-sdnq-standard.mp4
|
| 93 |
+
371afdc0bdc3281e88eed11ee3c94d4dd6592ff27265662ad4e1639fd9b191db benchmark/summary.json
|
| 94 |
+
8c94f2078b99415041f1c80bdb58959eaabcc63e68c137ebb030a1bbe594a316 lingbot_sdnq_runtime/__init__.py
|
| 95 |
+
4ce9e7219ce36b0ff660e9d6109727b2ec3b72ced4591e496dcb7a2c7431c777 lingbot_sdnq_runtime/__pycache__/__init__.cpython-312.pyc
|
| 96 |
+
ed29c8cadbb7b126732f49aa51379cf7d68450699957ad777001521fe290ccd2 lingbot_sdnq_runtime/__pycache__/__init__.cpython-314.pyc
|
| 97 |
+
92d1b6552667974106ef65d5595a673e66f85dcb68a9dac4f1406819c4c7d569 model_index.json
|
| 98 |
+
f1e022a5bc713b0bd516e9296451d1ef0384c96a0e85d7352549d0168f3c4098 model_index.json.bak_lingbot_video_diffusers
|
| 99 |
+
77f8f73bdd50919d92eacee3a9e62cee06959596f364283b8ff5ab0f2b128081 processor/README.md
|
| 100 |
+
6f8a6a55027e3da5160105556cda5dd69f6423f1c32645f6730d32de7773d0c4 processor/chat_template.json
|
| 101 |
+
edac7703329133edfc53e46ac0081835144c99d7eebf28b71c732694d435224d processor/config.json
|
| 102 |
+
f888421726665e8a84b738eed42a64875aed79de8be7daade851ac8bf4c0cef9 processor/configuration.json
|
| 103 |
+
8469742d1fce0de951c8909b26a2c0c0d8490837ce476efb114da9e0cefc4d44 processor/generation_config.json
|
| 104 |
+
599bab54075088774b1733fde865d5bd747cbcc7a547c5bc12610e874e26f5e3 processor/merges.txt
|
| 105 |
+
27225450ac9c6529872ee1924fcb0962ff5634834f817040f444118116f4e516 processor/preprocessor_config.json
|
| 106 |
+
a5d85b6dcc535e6b93115a9ef287e6132fdbf30270da6218194ba742261173c7 processor/tokenizer.json
|
| 107 |
+
c2da771801886ad9ae98181793ffd3dfb7f1af30f6f7c6a4e15d7dbba52e2399 processor/tokenizer_config.json
|
| 108 |
+
7768af27c1fafa9cc9011c1dc20067e03f8915e03b63504550e11d5066986d13 processor/video_preprocessor_config.json
|
| 109 |
+
ca10d7e9fb3ed18575dd1e277a2579c16d108e32f27439684afa0e10b1440910 processor/vocab.json
|
| 110 |
+
a7e0a5c16d9ce9acfa723aeb97e0ad75e095b75d6f5fe00b8c9a60bd555e52a9 prompts.json
|
| 111 |
+
12b99efd234716e35f35afd1a7a806082fc15c708af92ca43333b10ab86a0411 quantization_manifest.json
|
| 112 |
+
4f4c7ba0f76fc49d843eeb6c5a02c012cfafee68438f299f9aaa4eedd3f03bcc refiner/component.SHA256SUMS
|
| 113 |
+
2effac3d3a8c72dc3938d44b17b25e928b690717cb2e22a061df8aa3ed4a8c9a refiner/config.json
|
| 114 |
+
b21bc0b086b03cc9a0000cf7256a712dab2695ffcda7026b076c3abb99dd1eb3 refiner/coverage.json
|
| 115 |
+
f4fcfa25c57404a23e8da451e107b5533af5b579d907aa38b89f0d6a1a2828e7 refiner/diffusion_pytorch_model-00001-of-00005.safetensors
|
| 116 |
+
2489df8e6a45a7883fcff10b34faa0209f07d7b97e6c5681439fafe9d4041303 refiner/diffusion_pytorch_model-00002-of-00005.safetensors
|
| 117 |
+
e0cf002ed56e1e04140afb7137365ea5d8a3497ef7a4c55911b3db2fdf6545f5 refiner/diffusion_pytorch_model-00003-of-00005.safetensors
|
| 118 |
+
5bd87a652b9d279dbb5290a02c14ae6ce5987fdb0eb3caa30cea2980b2c63131 refiner/diffusion_pytorch_model-00004-of-00005.safetensors
|
| 119 |
+
0e5e9a038b797c68ea42398bbe425707c7b09414c352c6bfbd2923e7750f25ee refiner/diffusion_pytorch_model-00005-of-00005.safetensors
|
| 120 |
+
48d8cb723c47742c0e875083a3e699b8a4541d7fc1036b25a1a160746efe051d refiner/diffusion_pytorch_model.safetensors.index.json
|
| 121 |
+
e5ab2ee967920427f6897837cf55b798d10b0bce96635b92b315d747e8095c21 refiner/quantization_config.json
|
| 122 |
+
b49557484202c5499185e6af092f850d4b69788fddf2ccda57c06f5493887027 refiner/sdnq_experts.json
|
| 123 |
+
35a2b18534f58c028f8a3b23b5fe0e1f635c0cc4122d9f26cd497ff4b69797d2 runtime-requirements.txt
|
| 124 |
+
d040042bdb79aacdcafe80a2c4bd621271e3056823a84621c25ee8fd4fff594b samples/base/01_arctic_research_drone.mp4
|
| 125 |
+
bcb2caabb9e29427019ef3dcc7790a89910c1b570e65f85bac3f8a95dc1912db samples/base/02_macro_pastry_glaze.mp4
|
| 126 |
+
e91a616e88278e062635d12bd204679caf879dafc107cb066cf4155c3efe2486 samples/base/03_rain_market_lanterns.mp4
|
| 127 |
+
add7b1153f552850a2b095afa51998c4e3af4719f3d5c64425105f1af8c7458e samples/base/04_underwater_manta.mp4
|
| 128 |
+
7b6909a2aa49397f574338560cce4a3e47bb1930e55304f897523fa436fe7333 samples/base/05_factory_battery_cells.mp4
|
| 129 |
+
fb95a8d0cda0c3038e107a53eb3bc5b1df840a8568318c6f90bb4d0566a7674a samples/refiner/01_arctic_research_drone.mp4
|
| 130 |
+
b3bdfa0eaa7f4a3b9cbc3627f15f144b2416778f5160e3242bb2fc1ff284a7de samples/refiner-original/01_arctic_research_drone.mp4
|
| 131 |
+
74bdc2c22a985d24b6e053f17ab93139d048a8466b6fa77a98d52d518e8276ae scheduler/scheduler_config.json
|
| 132 |
+
4d057e930711803187bd17803f84e9ec43aff359a01f7a91abe115e0deecd08b scheduling_flow_unipc.py
|
| 133 |
+
c7753229ae8f0d881aa8d8151a55072c01569d3b53f5e15cd35533014a026e9b source_revisions.json
|
| 134 |
+
77f8f73bdd50919d92eacee3a9e62cee06959596f364283b8ff5ab0f2b128081 text_encoder/README.md
|
| 135 |
+
6f8a6a55027e3da5160105556cda5dd69f6423f1c32645f6730d32de7773d0c4 text_encoder/chat_template.json
|
| 136 |
+
edac7703329133edfc53e46ac0081835144c99d7eebf28b71c732694d435224d text_encoder/config.json
|
| 137 |
+
f888421726665e8a84b738eed42a64875aed79de8be7daade851ac8bf4c0cef9 text_encoder/configuration.json
|
| 138 |
+
8469742d1fce0de951c8909b26a2c0c0d8490837ce476efb114da9e0cefc4d44 text_encoder/generation_config.json
|
| 139 |
+
599bab54075088774b1733fde865d5bd747cbcc7a547c5bc12610e874e26f5e3 text_encoder/merges.txt
|
| 140 |
+
30a01a0556622645a3cce87b655bbbbbc1f170c196099f1b666c93202c3339a9 text_encoder/model-00001-of-00002.safetensors
|
| 141 |
+
046296a2a387efb43b0c997d5833c789604d168834f6e0d3064bf7bb13d002a6 text_encoder/model-00002-of-00002.safetensors
|
| 142 |
+
58a7841d7bff2548dd91577d216274a83cf1b500bc6a534b809d6c1b1707cf2b text_encoder/model.safetensors.index.json
|
| 143 |
+
27225450ac9c6529872ee1924fcb0962ff5634834f817040f444118116f4e516 text_encoder/preprocessor_config.json
|
| 144 |
+
a5d85b6dcc535e6b93115a9ef287e6132fdbf30270da6218194ba742261173c7 text_encoder/tokenizer.json
|
| 145 |
+
c2da771801886ad9ae98181793ffd3dfb7f1af30f6f7c6a4e15d7dbba52e2399 text_encoder/tokenizer_config.json
|
| 146 |
+
7768af27c1fafa9cc9011c1dc20067e03f8915e03b63504550e11d5066986d13 text_encoder/video_preprocessor_config.json
|
| 147 |
+
ca10d7e9fb3ed18575dd1e277a2579c16d108e32f27439684afa0e10b1440910 text_encoder/vocab.json
|
| 148 |
+
4eb27f30607d879d8d426b4f9f0e39f6b2642e5356a643b4b5a50dc17f6c1e83 transformer/component.SHA256SUMS
|
| 149 |
+
8f3e9ff1d5ec390da0d466f8ed8869042ae36e704a5fc7ecf8725ba689175da8 transformer/config.json
|
| 150 |
+
8bc6949365153e7e5779142a96de01ab23b8979f6c92a4d8ede7eea99e6ae6a5 transformer/coverage.json
|
| 151 |
+
380d327ee8172fd93869930380e705916558dd25341d957893f0f371b1203993 transformer/diffusion_pytorch_model-00001-of-00005.safetensors
|
| 152 |
+
7b3d48e0cbca336738276847ef5e44de42f03ada1cd762e56e9e86fcd93a49b6 transformer/diffusion_pytorch_model-00002-of-00005.safetensors
|
| 153 |
+
ed3461cb304ab56829a6cd34f379361578ef6eccabbdac87d028b39e4268ea25 transformer/diffusion_pytorch_model-00003-of-00005.safetensors
|
| 154 |
+
fc1c9feafa182672e3e33872a76575f53a4727f52e9df3bb81a36ab2b47be84a transformer/diffusion_pytorch_model-00004-of-00005.safetensors
|
| 155 |
+
37422bbc28d1d9b8793ab43ad584b22271db07fcf5957f3c66ee46a888bb750c transformer/diffusion_pytorch_model-00005-of-00005.safetensors
|
| 156 |
+
48d8cb723c47742c0e875083a3e699b8a4541d7fc1036b25a1a160746efe051d transformer/diffusion_pytorch_model.safetensors.index.json
|
| 157 |
+
f594ca9c58c2343b92e543cc7c37cba81c96aba36bc7410f1d645c82c80417bc transformer/quantization_config.json
|
| 158 |
+
b49557484202c5499185e6af092f850d4b69788fddf2ccda57c06f5493887027 transformer/sdnq_experts.json
|
| 159 |
+
f0c1cc1d7decb5badc384f54691746a27a9aeff49f7ebca974e583389342d527 vae/config.json
|
| 160 |
+
d6e524b3fffede1787a74e81b30976dce5400c4439ba64222168e607ed19e793 vae/diffusion_pytorch_model.safetensors
|
benchmark/audits/hub-weight-audit.json
ADDED
|
@@ -0,0 +1,178 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"schema_version": 1,
|
| 3 |
+
"checked_at_utc": "2026-07-10T08:14:46Z",
|
| 4 |
+
"audit_mode": "read_only",
|
| 5 |
+
"pod": {
|
| 6 |
+
"id": "dfencoyuz6u0kq",
|
| 7 |
+
"endpoint": "216.243.220.136:18411",
|
| 8 |
+
"source_root": "/workspace/lingbot-video-sdnq/quantized"
|
| 9 |
+
},
|
| 10 |
+
"method": {
|
| 11 |
+
"hub_api": "huggingface_hub HfApi.model_info(files_metadata=true)",
|
| 12 |
+
"comparison": "Hub LFS sha256 and byte size compared with the pod's immutable component.SHA256SUMS and stat size",
|
| 13 |
+
"visibility_changed": false
|
| 14 |
+
},
|
| 15 |
+
"repositories": {
|
| 16 |
+
"dense": {
|
| 17 |
+
"repo_id": "WaveCut/LingBot-Video-Dense-1.3B-SDNQ-uint4-static",
|
| 18 |
+
"revision": "ee9c3a469e4ad2a1f20054fe805444866594ba9b",
|
| 19 |
+
"private": true,
|
| 20 |
+
"gated": false,
|
| 21 |
+
"disabled": false,
|
| 22 |
+
"components": {
|
| 23 |
+
"transformer": {
|
| 24 |
+
"source_checksum_file": "/workspace/lingbot-video-sdnq/quantized/dense/transformer/component.SHA256SUMS",
|
| 25 |
+
"expected_file_count": 1,
|
| 26 |
+
"remote_file_count": 1,
|
| 27 |
+
"expected_bytes": 897102832,
|
| 28 |
+
"remote_bytes": 897102832,
|
| 29 |
+
"complete": true,
|
| 30 |
+
"files": [
|
| 31 |
+
{
|
| 32 |
+
"path": "transformer/diffusion_pytorch_model.safetensors",
|
| 33 |
+
"expected_sha256": "2556a4219537f51952a0289da3c0461860f67a20983d0df147d6b169fd89d627",
|
| 34 |
+
"remote_lfs_sha256": "2556a4219537f51952a0289da3c0461860f67a20983d0df147d6b169fd89d627",
|
| 35 |
+
"expected_bytes": 897102832,
|
| 36 |
+
"remote_lfs_bytes": 897102832,
|
| 37 |
+
"sha256_match": true,
|
| 38 |
+
"size_match": true
|
| 39 |
+
}
|
| 40 |
+
]
|
| 41 |
+
}
|
| 42 |
+
},
|
| 43 |
+
"weights_complete": true
|
| 44 |
+
},
|
| 45 |
+
"moe": {
|
| 46 |
+
"repo_id": "WaveCut/LingBot-Video-MoE-30B-A3B-SDNQ-uint4-static",
|
| 47 |
+
"revision": "206e412e861ef6fbdd89b61871d82523e19ede91",
|
| 48 |
+
"private": true,
|
| 49 |
+
"gated": false,
|
| 50 |
+
"disabled": false,
|
| 51 |
+
"components": {
|
| 52 |
+
"transformer": {
|
| 53 |
+
"source_checksum_file": "/workspace/lingbot-video-sdnq/quantized/moe/transformer/component.SHA256SUMS",
|
| 54 |
+
"expected_file_count": 5,
|
| 55 |
+
"remote_file_count": 5,
|
| 56 |
+
"expected_bytes": 18913365488,
|
| 57 |
+
"remote_bytes": 18913365488,
|
| 58 |
+
"complete": true,
|
| 59 |
+
"files": [
|
| 60 |
+
{
|
| 61 |
+
"path": "transformer/diffusion_pytorch_model-00001-of-00005.safetensors",
|
| 62 |
+
"expected_sha256": "380d327ee8172fd93869930380e705916558dd25341d957893f0f371b1203993",
|
| 63 |
+
"remote_lfs_sha256": "380d327ee8172fd93869930380e705916558dd25341d957893f0f371b1203993",
|
| 64 |
+
"expected_bytes": 3903708600,
|
| 65 |
+
"remote_lfs_bytes": 3903708600,
|
| 66 |
+
"sha256_match": true,
|
| 67 |
+
"size_match": true
|
| 68 |
+
},
|
| 69 |
+
{
|
| 70 |
+
"path": "transformer/diffusion_pytorch_model-00002-of-00005.safetensors",
|
| 71 |
+
"expected_sha256": "7b3d48e0cbca336738276847ef5e44de42f03ada1cd762e56e9e86fcd93a49b6",
|
| 72 |
+
"remote_lfs_sha256": "7b3d48e0cbca336738276847ef5e44de42f03ada1cd762e56e9e86fcd93a49b6",
|
| 73 |
+
"expected_bytes": 3907192200,
|
| 74 |
+
"remote_lfs_bytes": 3907192200,
|
| 75 |
+
"sha256_match": true,
|
| 76 |
+
"size_match": true
|
| 77 |
+
},
|
| 78 |
+
{
|
| 79 |
+
"path": "transformer/diffusion_pytorch_model-00003-of-00005.safetensors",
|
| 80 |
+
"expected_sha256": "ed3461cb304ab56829a6cd34f379361578ef6eccabbdac87d028b39e4268ea25",
|
| 81 |
+
"remote_lfs_sha256": "ed3461cb304ab56829a6cd34f379361578ef6eccabbdac87d028b39e4268ea25",
|
| 82 |
+
"expected_bytes": 3907192216,
|
| 83 |
+
"remote_lfs_bytes": 3907192216,
|
| 84 |
+
"sha256_match": true,
|
| 85 |
+
"size_match": true
|
| 86 |
+
},
|
| 87 |
+
{
|
| 88 |
+
"path": "transformer/diffusion_pytorch_model-00004-of-00005.safetensors",
|
| 89 |
+
"expected_sha256": "fc1c9feafa182672e3e33872a76575f53a4727f52e9df3bb81a36ab2b47be84a",
|
| 90 |
+
"remote_lfs_sha256": "fc1c9feafa182672e3e33872a76575f53a4727f52e9df3bb81a36ab2b47be84a",
|
| 91 |
+
"expected_bytes": 3907192216,
|
| 92 |
+
"remote_lfs_bytes": 3907192216,
|
| 93 |
+
"sha256_match": true,
|
| 94 |
+
"size_match": true
|
| 95 |
+
},
|
| 96 |
+
{
|
| 97 |
+
"path": "transformer/diffusion_pytorch_model-00005-of-00005.safetensors",
|
| 98 |
+
"expected_sha256": "37422bbc28d1d9b8793ab43ad584b22271db07fcf5957f3c66ee46a888bb750c",
|
| 99 |
+
"remote_lfs_sha256": "37422bbc28d1d9b8793ab43ad584b22271db07fcf5957f3c66ee46a888bb750c",
|
| 100 |
+
"expected_bytes": 3288080256,
|
| 101 |
+
"remote_lfs_bytes": 3288080256,
|
| 102 |
+
"sha256_match": true,
|
| 103 |
+
"size_match": true
|
| 104 |
+
}
|
| 105 |
+
]
|
| 106 |
+
},
|
| 107 |
+
"refiner": {
|
| 108 |
+
"source_checksum_file": "/workspace/lingbot-video-sdnq/quantized/moe/refiner/component.SHA256SUMS",
|
| 109 |
+
"expected_file_count": 5,
|
| 110 |
+
"remote_file_count": 5,
|
| 111 |
+
"expected_bytes": 18913365488,
|
| 112 |
+
"remote_bytes": 18913365488,
|
| 113 |
+
"complete": true,
|
| 114 |
+
"files": [
|
| 115 |
+
{
|
| 116 |
+
"path": "refiner/diffusion_pytorch_model-00001-of-00005.safetensors",
|
| 117 |
+
"expected_sha256": "f4fcfa25c57404a23e8da451e107b5533af5b579d907aa38b89f0d6a1a2828e7",
|
| 118 |
+
"remote_lfs_sha256": "f4fcfa25c57404a23e8da451e107b5533af5b579d907aa38b89f0d6a1a2828e7",
|
| 119 |
+
"expected_bytes": 3903708600,
|
| 120 |
+
"remote_lfs_bytes": 3903708600,
|
| 121 |
+
"sha256_match": true,
|
| 122 |
+
"size_match": true
|
| 123 |
+
},
|
| 124 |
+
{
|
| 125 |
+
"path": "refiner/diffusion_pytorch_model-00002-of-00005.safetensors",
|
| 126 |
+
"expected_sha256": "2489df8e6a45a7883fcff10b34faa0209f07d7b97e6c5681439fafe9d4041303",
|
| 127 |
+
"remote_lfs_sha256": "2489df8e6a45a7883fcff10b34faa0209f07d7b97e6c5681439fafe9d4041303",
|
| 128 |
+
"expected_bytes": 3907192200,
|
| 129 |
+
"remote_lfs_bytes": 3907192200,
|
| 130 |
+
"sha256_match": true,
|
| 131 |
+
"size_match": true
|
| 132 |
+
},
|
| 133 |
+
{
|
| 134 |
+
"path": "refiner/diffusion_pytorch_model-00003-of-00005.safetensors",
|
| 135 |
+
"expected_sha256": "e0cf002ed56e1e04140afb7137365ea5d8a3497ef7a4c55911b3db2fdf6545f5",
|
| 136 |
+
"remote_lfs_sha256": "e0cf002ed56e1e04140afb7137365ea5d8a3497ef7a4c55911b3db2fdf6545f5",
|
| 137 |
+
"expected_bytes": 3907192216,
|
| 138 |
+
"remote_lfs_bytes": 3907192216,
|
| 139 |
+
"sha256_match": true,
|
| 140 |
+
"size_match": true
|
| 141 |
+
},
|
| 142 |
+
{
|
| 143 |
+
"path": "refiner/diffusion_pytorch_model-00004-of-00005.safetensors",
|
| 144 |
+
"expected_sha256": "5bd87a652b9d279dbb5290a02c14ae6ce5987fdb0eb3caa30cea2980b2c63131",
|
| 145 |
+
"remote_lfs_sha256": "5bd87a652b9d279dbb5290a02c14ae6ce5987fdb0eb3caa30cea2980b2c63131",
|
| 146 |
+
"expected_bytes": 3907192216,
|
| 147 |
+
"remote_lfs_bytes": 3907192216,
|
| 148 |
+
"sha256_match": true,
|
| 149 |
+
"size_match": true
|
| 150 |
+
},
|
| 151 |
+
{
|
| 152 |
+
"path": "refiner/diffusion_pytorch_model-00005-of-00005.safetensors",
|
| 153 |
+
"expected_sha256": "0e5e9a038b797c68ea42398bbe425707c7b09414c352c6bfbd2923e7750f25ee",
|
| 154 |
+
"remote_lfs_sha256": "0e5e9a038b797c68ea42398bbe425707c7b09414c352c6bfbd2923e7750f25ee",
|
| 155 |
+
"expected_bytes": 3288080256,
|
| 156 |
+
"remote_lfs_bytes": 3288080256,
|
| 157 |
+
"sha256_match": true,
|
| 158 |
+
"size_match": true
|
| 159 |
+
}
|
| 160 |
+
]
|
| 161 |
+
}
|
| 162 |
+
},
|
| 163 |
+
"weights_complete": true
|
| 164 |
+
}
|
| 165 |
+
},
|
| 166 |
+
"summary": {
|
| 167 |
+
"all_expected_quantized_weight_files_present": true,
|
| 168 |
+
"all_lfs_sha256_match": true,
|
| 169 |
+
"all_lfs_sizes_match": true,
|
| 170 |
+
"verified_quantized_weight_files": 11,
|
| 171 |
+
"verified_quantized_weight_bytes": 38723833808,
|
| 172 |
+
"dense_transformer_complete": true,
|
| 173 |
+
"moe_transformer_complete": true,
|
| 174 |
+
"moe_refiner_complete": true,
|
| 175 |
+
"both_repositories_remain_private": true,
|
| 176 |
+
"acceptance": "pass"
|
| 177 |
+
}
|
| 178 |
+
}
|
benchmark/audits/numerical-smoke.json
ADDED
|
@@ -0,0 +1,23 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"compression_ratio": 3.2,
|
| 3 |
+
"expert_block": {
|
| 4 |
+
"finite": true,
|
| 5 |
+
"mae": 2.005113856284879e-05,
|
| 6 |
+
"max_abs": 0.00010204315185546875,
|
| 7 |
+
"missing_keys": [],
|
| 8 |
+
"rmse": 2.604380097182002e-05,
|
| 9 |
+
"roundtrip_max_abs": 0.0,
|
| 10 |
+
"unexpected_keys": []
|
| 11 |
+
},
|
| 12 |
+
"finite": true,
|
| 13 |
+
"mae": 0.0033151409588754177,
|
| 14 |
+
"max_abs": 0.0101318359375,
|
| 15 |
+
"original_bytes": 16384,
|
| 16 |
+
"rmse": 0.0039942022413015366,
|
| 17 |
+
"shape": [
|
| 18 |
+
4,
|
| 19 |
+
32,
|
| 20 |
+
64
|
| 21 |
+
],
|
| 22 |
+
"stored_bytes": 5120
|
| 23 |
+
}
|
benchmark/audits/source-audit.json
ADDED
|
@@ -0,0 +1,18 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"repo": "robbyant/lingbot-video-dense-1.3b",
|
| 4 |
+
"revision": "f9789a7d9b4772a47aba62d4eb5282ddefd1da21",
|
| 5 |
+
"expected_files": 35,
|
| 6 |
+
"missing": [],
|
| 7 |
+
"wrong_size": [],
|
| 8 |
+
"complete": true
|
| 9 |
+
},
|
| 10 |
+
{
|
| 11 |
+
"repo": "robbyant/lingbot-video-moe-30b-a3b",
|
| 12 |
+
"revision": "f2e538f64afe00cc4ae674db2aeb52e2945edfd5",
|
| 13 |
+
"expected_files": 63,
|
| 14 |
+
"missing": [],
|
| 15 |
+
"wrong_size": [],
|
| 16 |
+
"complete": true
|
| 17 |
+
}
|
| 18 |
+
]
|
benchmark/audits/video-encoding-audit.json
ADDED
|
@@ -0,0 +1,139 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"status": "verified",
|
| 3 |
+
"count": 22,
|
| 4 |
+
"max_midpoint_rgb_mae": 3.407521963119507,
|
| 5 |
+
"rows": [
|
| 6 |
+
{
|
| 7 |
+
"raw": "results/benchmark/dense/original/raw_frames/01_arctic_research_drone.npy",
|
| 8 |
+
"mp4": "results/benchmark/dense/original/videos/01_arctic_research_drone.mp4",
|
| 9 |
+
"frames": 73,
|
| 10 |
+
"midpoint_rgb_mae": 2.4514122009277344
|
| 11 |
+
},
|
| 12 |
+
{
|
| 13 |
+
"raw": "results/benchmark/dense/original/raw_frames/02_macro_pastry_glaze.npy",
|
| 14 |
+
"mp4": "results/benchmark/dense/original/videos/02_macro_pastry_glaze.mp4",
|
| 15 |
+
"frames": 73,
|
| 16 |
+
"midpoint_rgb_mae": 2.3606362342834473
|
| 17 |
+
},
|
| 18 |
+
{
|
| 19 |
+
"raw": "results/benchmark/dense/original/raw_frames/03_rain_market_lanterns.npy",
|
| 20 |
+
"mp4": "results/benchmark/dense/original/videos/03_rain_market_lanterns.mp4",
|
| 21 |
+
"frames": 73,
|
| 22 |
+
"midpoint_rgb_mae": 2.914302110671997
|
| 23 |
+
},
|
| 24 |
+
{
|
| 25 |
+
"raw": "results/benchmark/dense/original/raw_frames/04_underwater_manta.npy",
|
| 26 |
+
"mp4": "results/benchmark/dense/original/videos/04_underwater_manta.mp4",
|
| 27 |
+
"frames": 73,
|
| 28 |
+
"midpoint_rgb_mae": 3.407521963119507
|
| 29 |
+
},
|
| 30 |
+
{
|
| 31 |
+
"raw": "results/benchmark/dense/original/raw_frames/05_factory_battery_cells.npy",
|
| 32 |
+
"mp4": "results/benchmark/dense/original/videos/05_factory_battery_cells.mp4",
|
| 33 |
+
"frames": 73,
|
| 34 |
+
"midpoint_rgb_mae": 2.564425468444824
|
| 35 |
+
},
|
| 36 |
+
{
|
| 37 |
+
"raw": "results/benchmark/dense/sdnq/raw_frames/01_arctic_research_drone.npy",
|
| 38 |
+
"mp4": "results/benchmark/dense/sdnq/videos/01_arctic_research_drone.mp4",
|
| 39 |
+
"frames": 73,
|
| 40 |
+
"midpoint_rgb_mae": 2.5438008308410645
|
| 41 |
+
},
|
| 42 |
+
{
|
| 43 |
+
"raw": "results/benchmark/dense/sdnq/raw_frames/02_macro_pastry_glaze.npy",
|
| 44 |
+
"mp4": "results/benchmark/dense/sdnq/videos/02_macro_pastry_glaze.mp4",
|
| 45 |
+
"frames": 73,
|
| 46 |
+
"midpoint_rgb_mae": 2.2372729778289795
|
| 47 |
+
},
|
| 48 |
+
{
|
| 49 |
+
"raw": "results/benchmark/dense/sdnq/raw_frames/03_rain_market_lanterns.npy",
|
| 50 |
+
"mp4": "results/benchmark/dense/sdnq/videos/03_rain_market_lanterns.mp4",
|
| 51 |
+
"frames": 73,
|
| 52 |
+
"midpoint_rgb_mae": 3.0943760871887207
|
| 53 |
+
},
|
| 54 |
+
{
|
| 55 |
+
"raw": "results/benchmark/dense/sdnq/raw_frames/04_underwater_manta.npy",
|
| 56 |
+
"mp4": "results/benchmark/dense/sdnq/videos/04_underwater_manta.mp4",
|
| 57 |
+
"frames": 73,
|
| 58 |
+
"midpoint_rgb_mae": 3.361711263656616
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"raw": "results/benchmark/dense/sdnq/raw_frames/05_factory_battery_cells.npy",
|
| 62 |
+
"mp4": "results/benchmark/dense/sdnq/videos/05_factory_battery_cells.mp4",
|
| 63 |
+
"frames": 73,
|
| 64 |
+
"midpoint_rgb_mae": 2.557323455810547
|
| 65 |
+
},
|
| 66 |
+
{
|
| 67 |
+
"raw": "results/benchmark/moe/original/raw_frames/01_arctic_research_drone.npy",
|
| 68 |
+
"mp4": "results/benchmark/moe/original/videos/01_arctic_research_drone.mp4",
|
| 69 |
+
"frames": 73,
|
| 70 |
+
"midpoint_rgb_mae": 2.2264907360076904
|
| 71 |
+
},
|
| 72 |
+
{
|
| 73 |
+
"raw": "results/benchmark/moe/original/raw_frames/02_macro_pastry_glaze.npy",
|
| 74 |
+
"mp4": "results/benchmark/moe/original/videos/02_macro_pastry_glaze.mp4",
|
| 75 |
+
"frames": 73,
|
| 76 |
+
"midpoint_rgb_mae": 2.2814218997955322
|
| 77 |
+
},
|
| 78 |
+
{
|
| 79 |
+
"raw": "results/benchmark/moe/original/raw_frames/03_rain_market_lanterns.npy",
|
| 80 |
+
"mp4": "results/benchmark/moe/original/videos/03_rain_market_lanterns.mp4",
|
| 81 |
+
"frames": 73,
|
| 82 |
+
"midpoint_rgb_mae": 2.9506359100341797
|
| 83 |
+
},
|
| 84 |
+
{
|
| 85 |
+
"raw": "results/benchmark/moe/original/raw_frames/04_underwater_manta.npy",
|
| 86 |
+
"mp4": "results/benchmark/moe/original/videos/04_underwater_manta.mp4",
|
| 87 |
+
"frames": 73,
|
| 88 |
+
"midpoint_rgb_mae": 3.304417848587036
|
| 89 |
+
},
|
| 90 |
+
{
|
| 91 |
+
"raw": "results/benchmark/moe/original/raw_frames/05_factory_battery_cells.npy",
|
| 92 |
+
"mp4": "results/benchmark/moe/original/videos/05_factory_battery_cells.mp4",
|
| 93 |
+
"frames": 73,
|
| 94 |
+
"midpoint_rgb_mae": 2.621140480041504
|
| 95 |
+
},
|
| 96 |
+
{
|
| 97 |
+
"raw": "results/benchmark/moe/sdnq/raw_frames/01_arctic_research_drone.npy",
|
| 98 |
+
"mp4": "results/benchmark/moe/sdnq/videos/01_arctic_research_drone.mp4",
|
| 99 |
+
"frames": 73,
|
| 100 |
+
"midpoint_rgb_mae": 2.244572877883911
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"raw": "results/benchmark/moe/sdnq/raw_frames/02_macro_pastry_glaze.npy",
|
| 104 |
+
"mp4": "results/benchmark/moe/sdnq/videos/02_macro_pastry_glaze.mp4",
|
| 105 |
+
"frames": 73,
|
| 106 |
+
"midpoint_rgb_mae": 2.295715570449829
|
| 107 |
+
},
|
| 108 |
+
{
|
| 109 |
+
"raw": "results/benchmark/moe/sdnq/raw_frames/03_rain_market_lanterns.npy",
|
| 110 |
+
"mp4": "results/benchmark/moe/sdnq/videos/03_rain_market_lanterns.mp4",
|
| 111 |
+
"frames": 73,
|
| 112 |
+
"midpoint_rgb_mae": 2.7779204845428467
|
| 113 |
+
},
|
| 114 |
+
{
|
| 115 |
+
"raw": "results/benchmark/moe/sdnq/raw_frames/04_underwater_manta.npy",
|
| 116 |
+
"mp4": "results/benchmark/moe/sdnq/videos/04_underwater_manta.mp4",
|
| 117 |
+
"frames": 73,
|
| 118 |
+
"midpoint_rgb_mae": 2.977720260620117
|
| 119 |
+
},
|
| 120 |
+
{
|
| 121 |
+
"raw": "results/benchmark/moe/sdnq/raw_frames/05_factory_battery_cells.npy",
|
| 122 |
+
"mp4": "results/benchmark/moe/sdnq/videos/05_factory_battery_cells.mp4",
|
| 123 |
+
"frames": 73,
|
| 124 |
+
"midpoint_rgb_mae": 1.429641604423523
|
| 125 |
+
},
|
| 126 |
+
{
|
| 127 |
+
"raw": "results/benchmark/moe_refiner/original/raw_frames/01_arctic_research_drone.npy",
|
| 128 |
+
"mp4": "results/benchmark/moe_refiner/original/videos/01_arctic_research_drone.mp4",
|
| 129 |
+
"frames": 73,
|
| 130 |
+
"midpoint_rgb_mae": 2.014193058013916
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"raw": "results/benchmark/moe_refiner/sdnq/raw_frames/01_arctic_research_drone.npy",
|
| 134 |
+
"mp4": "results/benchmark/moe_refiner/sdnq/videos/01_arctic_research_drone.mp4",
|
| 135 |
+
"frames": 73,
|
| 136 |
+
"midpoint_rgb_mae": 2.048431873321533
|
| 137 |
+
}
|
| 138 |
+
]
|
| 139 |
+
}
|
benchmark/audits/video-encoding-repair.json
ADDED
|
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"outputs": [
|
| 3 |
+
{
|
| 4 |
+
"directory": "/workspace/lingbot-video-sdnq/results/benchmark/dense/original",
|
| 5 |
+
"videos": 5
|
| 6 |
+
},
|
| 7 |
+
{
|
| 8 |
+
"directory": "/workspace/lingbot-video-sdnq/results/benchmark/dense/sdnq",
|
| 9 |
+
"videos": 5
|
| 10 |
+
},
|
| 11 |
+
{
|
| 12 |
+
"directory": "/workspace/lingbot-video-sdnq/results/benchmark/moe/original",
|
| 13 |
+
"videos": 5
|
| 14 |
+
},
|
| 15 |
+
{
|
| 16 |
+
"directory": "/workspace/lingbot-video-sdnq/results/benchmark/moe/sdnq",
|
| 17 |
+
"videos": 5
|
| 18 |
+
},
|
| 19 |
+
{
|
| 20 |
+
"directory": "/workspace/lingbot-video-sdnq/results/benchmark/moe_refiner/original",
|
| 21 |
+
"videos": 1
|
| 22 |
+
},
|
| 23 |
+
{
|
| 24 |
+
"directory": "/workspace/lingbot-video-sdnq/results/benchmark/moe_refiner/sdnq",
|
| 25 |
+
"videos": 1
|
| 26 |
+
}
|
| 27 |
+
],
|
| 28 |
+
"status": "verified"
|
| 29 |
+
}
|
benchmark/base/original/command.json
ADDED
|
@@ -0,0 +1,21 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"argv": [
|
| 3 |
+
"benchmark_base.py",
|
| 4 |
+
"--model-key",
|
| 5 |
+
"moe",
|
| 6 |
+
"--variant",
|
| 7 |
+
"original",
|
| 8 |
+
"--model",
|
| 9 |
+
"models/moe",
|
| 10 |
+
"--source-repo-id",
|
| 11 |
+
"robbyant/lingbot-video-moe-30b-a3b",
|
| 12 |
+
"--source-revision",
|
| 13 |
+
"f2e538f64afe00cc4ae674db2aeb52e2945edfd5",
|
| 14 |
+
"--output",
|
| 15 |
+
"results/benchmark/moe/original",
|
| 16 |
+
"--prompts",
|
| 17 |
+
"inputs/prompts"
|
| 18 |
+
],
|
| 19 |
+
"finished_utc": "2026-07-10T08:39:38Z",
|
| 20 |
+
"return_code": 0
|
| 21 |
+
}
|
benchmark/base/original/metrics.csv
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
model_key,variant,prompt_id,seed,generation_seconds,seconds_per_frame,torch_peak_allocated_mb,torch_peak_reserved_mb,peak_gpu_memory_used_mb,peak_gpu_utilization_percent,peak_process_rss_bytes,peak_system_used_bytes,output_size_bytes,output_sha256
|
| 2 |
+
moe,original,01_arctic_research_drone,4201,140.9023737310199,1.9301695031646562,112469.38330078125,124524.0,125646.0,100.0,2700222464.0,80392216576.0,366044,7a548d09dd53755c1fdd7e759aae7cf92fd57d555bdc13f28023ede4ffe5f95e
|
| 3 |
+
moe,original,02_macro_pastry_glaze,4202,139.39877224492375,1.9095722225332021,112468.59130859375,124524.0,126274.0,100.0,3164041216.0,90717732864.0,396167,0073c4f94ed8fc0b9bfaa93966ac627b8a66f31efe9547a78bc6934ee2ecfd82
|
| 4 |
+
moe,original,03_rain_market_lanterns,4203,139.6812952979235,1.9134424013414177,112469.681640625,124524.0,125646.0,100.0,3218972672.0,91209596928.0,430040,753f2016262fbd6a3644b28ad4f6243e15e8b55ef9474616a9c7d636f474090b
|
| 5 |
+
moe,original,04_underwater_manta,4204,139.6351271129679,1.9128099604516149,112469.43212890625,124524.0,125646.0,100.0,3070435328.0,91173777408.0,359114,dcd3ffe1e7dd1f72f86e256d0db51acfb6661b880bfb0b2a5b798c9609cae24a
|
| 6 |
+
moe,original,05_factory_battery_cells,4205,139.8160271209199,1.9152880427523273,112470.375,124524.0,125646.0,100.0,3052527616.0,92903628800.0,324116,8a08377ecb3877e3ab261ca1840450dfa28a18ccc6cc59679002da1ff7750b49
|
benchmark/base/original/metrics.jsonl
ADDED
|
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"batch_cfg": false, "cold_generation": true, "ffprobe": {"programs": [], "streams": [{"codec_name": "h264", "duration": "3.041667", "height": 480, "nb_frames": "73", "pix_fmt": "yuv420p", "r_frame_rate": "24/1", "width": 832}]}, "fps": 24, "generation_seconds": 140.9023737310199, "guidance_scale": 3.0, "height": 480, "model_key": "moe", "negative_prompt": "{\"universal_negative\": {\"visual_quality\": [\"low quality\", \"worst quality\", \"blurry\", \"pixelated\", \"jpeg artifacts\", \"low resolution\", \"unstable color\", \"color flicker\", \"underexposed\", \"overexposed\", \"invisible subject\", \"subject hidden in darkness\"], \"artistic_style\": [\"painting\", \"illustration\", \"drawing\", \"cartoon\", \"3d render\", \"cgi\", \"sketch\", \"digital art\"], \"composition_and_content\": [\"text\", \"watermark\", \"signature\", \"logo\", \"subtitles\", \"pillarboxed\", \"side bars\", \"portrait image in landscape frame\"], \"temporal_and_motion_stability\": [\"flickering\", \"jittery\", \"motion blur\", \"temporal inconsistency\", \"warping\", \"morphing\", \"incoherent motion\", \"unnatural movement\", \"static object with sudden jump\", \"frame-to-frame inconsistency\"], \"material_and_structure\": [\"plastic-like glass\", \"unrealistic texture\", \"deformed bottle\", \"liquid freezing improperly\", \"distorted reflections\"]}}", "null_cond_clone_zero": false, "num_frames": 73, "num_inference_steps": 40, "nvidia_smi_after": {"memory.total": "183359", "memory.used": "125646", "name": "NVIDIA B200", "power.draw": "260.66", "temperature.gpu": "43", "timestamp": "2026/07/10 08:30:06.448", "utilization.gpu": "0"}, "output_file": "results/benchmark/moe/original/videos/01_arctic_research_drone.mp4", "output_sha256": "7a548d09dd53755c1fdd7e759aae7cf92fd57d555bdc13f28023ede4ffe5f95e", "output_size_bytes": 366044, "peak_disk_used_bytes": 188024430592.0, "peak_gpu_memory_used_mb": 125646.0, "peak_gpu_power_watts": 977.6, "peak_gpu_utilization_percent": 100.0, "peak_process_rss_bytes": 2700222464.0, "peak_system_used_bytes": 80392216576.0, "prompt": "{\"comprehensive_description\":{\"scene_content_description\":\"A cinematic dusk shot of an Arctic research outpost on a wind-polished ice shelf under green aurora ribbons. A compact orange rover with a roof-mounted lidar mast drives slowly between blue instrument crates and slender weather antennas. Fine snow dust trails behind the tires, and warm light glows from the station windows against the cold cyan landscape. The atmosphere is realistic, high-detail documentary science footage with crisp snow texture, vapor in the air, and physically plausible shadows.\",\"camera_movement_description\":\"The camera begins as a low, smooth drone tracking shot behind the rover, then arcs gently left to reveal the glowing station and aurora-filled sky while keeping the rover dominant in frame.\"},\"camera_info\":{\"color\":\"Cold cyan with warm orange accents\",\"frame_size\":\"Wide\",\"shot_type_angle\":\"Low angle tracking shot\",\"lens_size\":\"Wide lens\",\"composition\":\"Rover in lower center, research station in the background, aurora above\",\"lighting\":\"Soft twilight with window glow\",\"lighting_type\":\"Natural dusk and practical interior lights\"},\"world_knowledge\":[],\"prominent_elements\":[{\"name\":\"orange autonomous research rover\",\"description\":\"A rugged six-wheeled scientific rover with orange body panels, black tires, a lidar mast, and small blinking status lights.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.7s]\",\"action\":\"drives forward slowly across packed snow, tires compressing the surface\"},{\"timestamp\":\"[0.7s - 1.5s]\",\"action\":\"turns slightly left as the lidar mast rotates\"},{\"timestamp\":\"[1.5s - 2.0s]\",\"action\":\"continues toward the lit station while fine snow trails behind\"}],\"location\":\"lower center of the frame\",\"relative_size\":\"dominant\",\"shape_and_color\":\"low rectangular orange body with black wheels and grey sensor mast\",\"texture\":\"matte painted metal, rubber tires, frost on edges\",\"appearance_details\":\"roof lidar, small antennas, narrow headlights, compact cargo rack\",\"relationship\":\"the main moving subject, traveling toward the research outpost\",\"orientation\":\"moving away from the camera and slightly left\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"Arctic research station\",\"description\":\"A modular polar station made of connected white containers with warm yellow light in the windows.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"remains stationary as wind-blown snow passes in front of it\"}],\"location\":\"middle background\",\"relative_size\":\"large\",\"shape_and_color\":\"rectangular white modules with dark seams and glowing windows\",\"texture\":\"frosted metal panels and glass\",\"appearance_details\":\"small stairs, railings, antennas, blue equipment crates nearby\",\"relationship\":\"destination of the rover\",\"orientation\":\"angled three-quarter view\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"aurora ribbons\",\"description\":\"Green aurora curtains stretching across the darkening polar sky.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"shimmer slowly and drift laterally across the sky\"}],\"location\":\"upper half of the frame\",\"relative_size\":\"large\",\"shape_and_color\":\"soft green luminous ribbons\",\"texture\":\"translucent atmospheric glow\",\"appearance_details\":\"layered folds and faint stars behind\",\"relationship\":\"dominates the sky above the station\",\"orientation\":\"horizontal sweeping arcs\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":true,\"number_of_objects\":\"many luminous bands\"}]}", "prompt_file": "01_arctic_research_drone.json", "prompt_id": "01_arctic_research_drone", "raw_frames_file": "results/benchmark/moe/original/raw_frames/01_arctic_research_drone.npy", "raw_frames_sha256": "1c1cad878f2c8f39889f0acb71d6ebaf1b12222ff43ddcd8419e54e9cf00d590", "raw_frames_size_bytes": 87459968, "sample_count": 488, "sample_interval_seconds": 0.25, "seconds_per_frame": 1.9301695031646562, "seed": 4201, "shift": 3.0, "torch_allocated_after_mb": 103497.9833984375, "torch_peak_allocated_mb": 112469.38330078125, "torch_peak_reserved_mb": 124524.0, "torch_reserved_after_mb": 124524.0, "variant": "original", "width": 832}
|
| 2 |
+
{"batch_cfg": false, "cold_generation": false, "ffprobe": {"programs": [], "streams": [{"codec_name": "h264", "duration": "3.041667", "height": 480, "nb_frames": "73", "pix_fmt": "yuv420p", "r_frame_rate": "24/1", "width": 832}]}, "fps": 24, "generation_seconds": 139.39877224492375, "guidance_scale": 3.0, "height": 480, "model_key": "moe", "negative_prompt": "{\"universal_negative\": {\"visual_quality\": [\"low quality\", \"worst quality\", \"blurry\", \"pixelated\", \"jpeg artifacts\", \"low resolution\", \"unstable color\", \"color flicker\", \"underexposed\", \"overexposed\", \"invisible subject\", \"subject hidden in darkness\"], \"artistic_style\": [\"painting\", \"illustration\", \"drawing\", \"cartoon\", \"3d render\", \"cgi\", \"sketch\", \"digital art\"], \"composition_and_content\": [\"text\", \"watermark\", \"signature\", \"logo\", \"subtitles\", \"pillarboxed\", \"side bars\", \"portrait image in landscape frame\"], \"temporal_and_motion_stability\": [\"flickering\", \"jittery\", \"motion blur\", \"temporal inconsistency\", \"warping\", \"morphing\", \"incoherent motion\", \"unnatural movement\", \"static object with sudden jump\", \"frame-to-frame inconsistency\"], \"material_and_structure\": [\"plastic-like glass\", \"unrealistic texture\", \"deformed bottle\", \"liquid freezing improperly\", \"distorted reflections\"]}}", "null_cond_clone_zero": false, "num_frames": 73, "num_inference_steps": 40, "nvidia_smi_after": {"memory.total": "183359", "memory.used": "125646", "name": "NVIDIA B200", "power.draw": "260.70", "temperature.gpu": "43", "timestamp": "2026/07/10 08:32:29.979", "utilization.gpu": "0"}, "output_file": "results/benchmark/moe/original/videos/02_macro_pastry_glaze.mp4", "output_sha256": "0073c4f94ed8fc0b9bfaa93966ac627b8a66f31efe9547a78bc6934ee2ecfd82", "output_size_bytes": 396167, "peak_disk_used_bytes": 188112715776.0, "peak_gpu_memory_used_mb": 126274.0, "peak_gpu_power_watts": 969.78, "peak_gpu_utilization_percent": 100.0, "peak_process_rss_bytes": 3164041216.0, "peak_system_used_bytes": 90717732864.0, "prompt": "{\"comprehensive_description\":{\"scene_content_description\":\"An extreme macro food cinematography shot in a professional pastry kitchen. A glossy dark chocolate dome dessert sits on a rotating metal turntable. A thin stream of amber caramel glaze pours from above, flowing over the curved surface in slow ribbons and revealing tiny reflections from softbox lights. Crushed pistachio dust and edible gold flakes sit around the base on a matte black plate. The look is photorealistic, tactile, high-speed macro video with shallow depth of field and no text.\",\"camera_movement_description\":\"The camera holds a close macro angle with a subtle push-in while the dessert rotates clockwise on the turntable.\"},\"camera_info\":{\"color\":\"Deep brown, amber, green pistachio, gold highlights\",\"frame_size\":\"Extreme close up\",\"shot_type_angle\":\"Slight high angle\",\"lens_size\":\"Macro lens\",\"composition\":\"Dessert centered with glaze entering from top\",\"lighting\":\"Large softbox reflections\",\"lighting_type\":\"Controlled studio kitchen lighting\"},\"world_knowledge\":[],\"prominent_elements\":[{\"name\":\"chocolate dome dessert\",\"description\":\"A small mirror-glazed chocolate dome on a black plate, surrounded by pistachio dust and gold flakes.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"rotates slowly clockwise on a metal turntable\"}],\"location\":\"center of the frame\",\"relative_size\":\"dominant\",\"shape_and_color\":\"hemispherical dark chocolate dome on round black plate\",\"texture\":\"mirror-gloss surface with smooth curved reflections\",\"appearance_details\":\"tiny gold flakes, green pistachio crumbs, clean pastry plating\",\"relationship\":\"receives the caramel glaze\",\"orientation\":\"upright on the turntable\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"caramel glaze stream\",\"description\":\"A viscous amber caramel stream pouring from above and spreading over the dessert.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.8s]\",\"action\":\"falls as a narrow glossy stream onto the crown of the dome\"},{\"timestamp\":\"[0.8s - 1.6s]\",\"action\":\"forms slow ribbons that slide down the curved chocolate surface\"},{\"timestamp\":\"[1.6s - 2.0s]\",\"action\":\"collects at the base in a thin shiny ring\"}],\"location\":\"entering from top center and flowing over the dessert\",\"relative_size\":\"medium\",\"shape_and_color\":\"thin amber liquid ribbons\",\"texture\":\"viscous, glossy, translucent caramel\",\"appearance_details\":\"bright highlights and tiny bubbles\",\"relationship\":\"coats the chocolate dome\",\"orientation\":\"vertical stream becoming downward trails\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"}]}", "prompt_file": "02_macro_pastry_glaze.json", "prompt_id": "02_macro_pastry_glaze", "raw_frames_file": "results/benchmark/moe/original/raw_frames/02_macro_pastry_glaze.npy", "raw_frames_sha256": "c96f6b8795b970785ca2c50c383ab9e6055b18c05496bbe0873aa67dbea063ec", "raw_frames_size_bytes": 87459968, "sample_count": 463, "sample_interval_seconds": 0.25, "seconds_per_frame": 1.9095722225332021, "seed": 4202, "shift": 3.0, "torch_allocated_after_mb": 103497.9833984375, "torch_peak_allocated_mb": 112468.59130859375, "torch_peak_reserved_mb": 124524.0, "torch_reserved_after_mb": 124524.0, "variant": "original", "width": 832}
|
| 3 |
+
{"batch_cfg": false, "cold_generation": false, "ffprobe": {"programs": [], "streams": [{"codec_name": "h264", "duration": "3.041667", "height": 480, "nb_frames": "73", "pix_fmt": "yuv420p", "r_frame_rate": "24/1", "width": 832}]}, "fps": 24, "generation_seconds": 139.6812952979235, "guidance_scale": 3.0, "height": 480, "model_key": "moe", "negative_prompt": "{\"universal_negative\": {\"visual_quality\": [\"low quality\", \"worst quality\", \"blurry\", \"pixelated\", \"jpeg artifacts\", \"low resolution\", \"unstable color\", \"color flicker\", \"underexposed\", \"overexposed\", \"invisible subject\", \"subject hidden in darkness\"], \"artistic_style\": [\"painting\", \"illustration\", \"drawing\", \"cartoon\", \"3d render\", \"cgi\", \"sketch\", \"digital art\"], \"composition_and_content\": [\"text\", \"watermark\", \"signature\", \"logo\", \"subtitles\", \"pillarboxed\", \"side bars\", \"portrait image in landscape frame\"], \"temporal_and_motion_stability\": [\"flickering\", \"jittery\", \"motion blur\", \"temporal inconsistency\", \"warping\", \"morphing\", \"incoherent motion\", \"unnatural movement\", \"static object with sudden jump\", \"frame-to-frame inconsistency\"], \"material_and_structure\": [\"plastic-like glass\", \"unrealistic texture\", \"deformed bottle\", \"liquid freezing improperly\", \"distorted reflections\"]}}", "null_cond_clone_zero": false, "num_frames": 73, "num_inference_steps": 40, "nvidia_smi_after": {"memory.total": "183359", "memory.used": "125646", "name": "NVIDIA B200", "power.draw": "260.53", "temperature.gpu": "43", "timestamp": "2026/07/10 08:34:52.790", "utilization.gpu": "0"}, "output_file": "results/benchmark/moe/original/videos/03_rain_market_lanterns.mp4", "output_sha256": "753f2016262fbd6a3644b28ad4f6243e15e8b55ef9474616a9c7d636f474090b", "output_size_bytes": 430040, "peak_disk_used_bytes": 188200804352.0, "peak_gpu_memory_used_mb": 125646.0, "peak_gpu_power_watts": 972.57, "peak_gpu_utilization_percent": 100.0, "peak_process_rss_bytes": 3218972672.0, "peak_system_used_bytes": 91209596928.0, "prompt": "{\"comprehensive_description\":{\"scene_content_description\":\"A lively rainy night market in a narrow old-city street, filmed like naturalistic cinema. Red paper lanterns hang from wooden stalls, steam rises from bowls of soup, and wet cobblestones mirror orange and teal reflections. A vendor in a dark apron slides a steaming bowl across a counter to a traveler in a blue raincoat. Umbrellas pass in the foreground, partially occluding the scene for a moment. The mood is warm, humid, detailed, and grounded, with no readable signs or subtitles.\",\"camera_movement_description\":\"The camera performs a slow lateral dolly from left to right at counter height, with foreground umbrellas briefly crossing the lens.\"},\"camera_info\":{\"color\":\"Warm lantern orange, teal rain reflections, red accents\",\"frame_size\":\"Medium wide\",\"shot_type_angle\":\"Eye level\",\"lens_size\":\"Medium lens\",\"composition\":\"Vendor stall centered, traveler on right, passing umbrellas in foreground\",\"lighting\":\"Lantern light mixed with rain reflections\",\"lighting_type\":\"Night practical lighting\"},\"world_knowledge\":[],\"prominent_elements\":[{\"name\":\"food vendor\",\"description\":\"A middle-aged street vendor wearing a dark apron and working behind a steaming wooden food stall.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.8s]\",\"action\":\"lifts a white ceramic bowl from a steaming pot\"},{\"timestamp\":\"[0.8s - 1.6s]\",\"action\":\"slides the bowl carefully across the counter\"},{\"timestamp\":\"[1.6s - 2.0s]\",\"action\":\"wipes condensation from the counter with a cloth\"}],\"location\":\"center-left behind the stall counter\",\"relative_size\":\"large\",\"shape_and_color\":\"human figure in dark apron under warm lantern light\",\"texture\":\"wet fabric, skin highlights, wood grain around the stall\",\"appearance_details\":\"rolled sleeves, focused expression, steam around hands\",\"relationship\":\"serves food to the traveler\",\"orientation\":\"facing right toward the customer\",\"pose\":\"leaning slightly forward over the counter\",\"expression\":\"focused and calm\",\"clothing\":\"dark apron over a muted shirt\",\"gender\":\"\",\"skin_tone_and_texture\":\"natural skin with rain-lit highlights\"},{\"name\":\"traveler in blue raincoat\",\"description\":\"A traveler wearing a blue hooded raincoat, waiting at the stall with wet sleeves and a folded umbrella.\",\"actions\":[{\"timestamp\":\"[0.0s - 1.2s]\",\"action\":\"waits with hands near the counter, watching the bowl\"},{\"timestamp\":\"[1.2s - 2.0s]\",\"action\":\"reaches forward to receive the steaming bowl\"}],\"location\":\"right side of the frame\",\"relative_size\":\"medium\",\"shape_and_color\":\"blue hooded raincoat silhouette\",\"texture\":\"water beads on waterproof fabric\",\"appearance_details\":\"hood up, folded umbrella tucked under one arm\",\"relationship\":\"customer receiving food from the vendor\",\"orientation\":\"facing left toward the stall\",\"pose\":\"standing close to the counter\",\"expression\":\"expectant and tired\",\"clothing\":\"blue raincoat\",\"gender\":\"\",\"skin_tone_and_texture\":\"\"},{\"name\":\"rain and lantern reflections\",\"description\":\"Falling rain, steam, and glowing reflections on wet cobblestones.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"rain falls continuously while reflections ripple on the street\"}],\"location\":\"foreground and background throughout the frame\",\"relative_size\":\"large\",\"shape_and_color\":\"thin rain streaks, orange and teal reflected pools\",\"texture\":\"wet stone, vapor, glossy water\",\"appearance_details\":\"soft steam plumes, umbrellas passing close to lens\",\"relationship\":\"sets the atmosphere around the market\",\"orientation\":\"vertical rainfall and horizontal street reflections\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":true,\"number_of_objects\":\"many raindrops and reflections\"}]}", "prompt_file": "03_rain_market_lanterns.json", "prompt_id": "03_rain_market_lanterns", "raw_frames_file": "results/benchmark/moe/original/raw_frames/03_rain_market_lanterns.npy", "raw_frames_sha256": "8b0391e7276b04d2d802ce81b1f1a32d5e4cd0c976f55d2466daaa3a6c838a63", "raw_frames_size_bytes": 87459968, "sample_count": 486, "sample_interval_seconds": 0.25, "seconds_per_frame": 1.9134424013414177, "seed": 4203, "shift": 3.0, "torch_allocated_after_mb": 103497.9833984375, "torch_peak_allocated_mb": 112469.681640625, "torch_peak_reserved_mb": 124524.0, "torch_reserved_after_mb": 124524.0, "variant": "original", "width": 832}
|
| 4 |
+
{"batch_cfg": false, "cold_generation": false, "ffprobe": {"programs": [], "streams": [{"codec_name": "h264", "duration": "3.041667", "height": 480, "nb_frames": "73", "pix_fmt": "yuv420p", "r_frame_rate": "24/1", "width": 832}]}, "fps": 24, "generation_seconds": 139.6351271129679, "guidance_scale": 3.0, "height": 480, "model_key": "moe", "negative_prompt": "{\"universal_negative\": {\"visual_quality\": [\"low quality\", \"worst quality\", \"blurry\", \"pixelated\", \"jpeg artifacts\", \"low resolution\", \"unstable color\", \"color flicker\", \"underexposed\", \"overexposed\", \"invisible subject\", \"subject hidden in darkness\"], \"artistic_style\": [\"painting\", \"illustration\", \"drawing\", \"cartoon\", \"3d render\", \"cgi\", \"sketch\", \"digital art\"], \"composition_and_content\": [\"text\", \"watermark\", \"signature\", \"logo\", \"subtitles\", \"pillarboxed\", \"side bars\", \"portrait image in landscape frame\"], \"temporal_and_motion_stability\": [\"flickering\", \"jittery\", \"motion blur\", \"temporal inconsistency\", \"warping\", \"morphing\", \"incoherent motion\", \"unnatural movement\", \"static object with sudden jump\", \"frame-to-frame inconsistency\"], \"material_and_structure\": [\"plastic-like glass\", \"unrealistic texture\", \"deformed bottle\", \"liquid freezing improperly\", \"distorted reflections\"]}}", "null_cond_clone_zero": false, "num_frames": 73, "num_inference_steps": 40, "nvidia_smi_after": {"memory.total": "183359", "memory.used": "125646", "name": "NVIDIA B200", "power.draw": "261.37", "temperature.gpu": "43", "timestamp": "2026/07/10 08:37:15.363", "utilization.gpu": "0"}, "output_file": "results/benchmark/moe/original/videos/04_underwater_manta.mp4", "output_sha256": "dcd3ffe1e7dd1f72f86e256d0db51acfb6661b880bfb0b2a5b798c9609cae24a", "output_size_bytes": 359114, "peak_disk_used_bytes": 188289716224.0, "peak_gpu_memory_used_mb": 125646.0, "peak_gpu_power_watts": 978.59, "peak_gpu_utilization_percent": 100.0, "peak_process_rss_bytes": 3070435328.0, "peak_system_used_bytes": 91173777408.0, "prompt": "{\"comprehensive_description\":{\"scene_content_description\":\"A tranquil underwater wildlife scene in clear tropical water. A large manta ray glides above a colorful coral garden while a scuba diver hovers several meters behind, holding a small camera rig. Sunbeams ripple through the surface, schools of tiny silver fish split and rejoin around the manta, and suspended particles drift slowly. The image should feel like realistic underwater documentary footage with natural motion, soft blue-green color, and no fantasy elements.\",\"camera_movement_description\":\"The camera tracks smoothly beside the manta ray from a slightly lower angle, drifting forward with the current.\"},\"camera_info\":{\"color\":\"Blue-green water with coral reds and yellows\",\"frame_size\":\"Wide\",\"shot_type_angle\":\"Slight low angle\",\"lens_size\":\"Wide underwater lens\",\"composition\":\"Manta ray crossing the upper center, coral below, diver in background\",\"lighting\":\"Dappled sunlight\",\"lighting_type\":\"Natural underwater daylight\"},\"world_knowledge\":[],\"prominent_elements\":[{\"name\":\"manta ray\",\"description\":\"A large manta ray with broad triangular fins, dark top surface, and pale underside.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.8s]\",\"action\":\"glides from left to right with slow wing-like fin movement\"},{\"timestamp\":\"[0.8s - 1.5s]\",\"action\":\"tilts slightly upward as sunbeams cross its back\"},{\"timestamp\":\"[1.5s - 2.0s]\",\"action\":\"continues forward while small fish scatter around it\"}],\"location\":\"upper center moving toward right\",\"relative_size\":\"dominant\",\"shape_and_color\":\"wide diamond-like silhouette, dark grey top, pale underside\",\"texture\":\"smooth skin with subtle mottling\",\"appearance_details\":\"cephalic fins near mouth, long tail trailing behind\",\"relationship\":\"main wildlife subject of the video\",\"orientation\":\"moving left to right and slightly upward\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"scuba diver\",\"description\":\"A diver in black wetsuit and fins, carrying a small underwater camera rig.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"hovers calmly in the background, exhaling small bubble streams\"}],\"location\":\"mid-background behind the manta ray\",\"relative_size\":\"small\",\"shape_and_color\":\"human figure in black wetsuit with silver air tank\",\"texture\":\"neoprene suit, metal tank, glass mask\",\"appearance_details\":\"fins, mask, regulator, compact camera lights\",\"relationship\":\"observes the manta ray without touching it\",\"orientation\":\"facing the manta ray\",\"pose\":\"horizontal hover\",\"expression\":\"\",\"clothing\":\"black scuba gear\",\"gender\":\"\",\"skin_tone_and_texture\":\"\"},{\"name\":\"coral reef and fish\",\"description\":\"A colorful coral garden with small silver fish schooling above it.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"fish swirl and split around the manta while coral remains fixed\"}],\"location\":\"lower half of the frame\",\"relative_size\":\"large\",\"shape_and_color\":\"branching corals in red, yellow, and muted purple; silver fish clusters\",\"texture\":\"rough coral, shimmering fish scales\",\"appearance_details\":\"sand patches, sea fans, drifting particles\",\"relationship\":\"environment below the manta ray\",\"orientation\":\"reef spreads horizontally across the bottom\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":true,\"number_of_objects\":\"many fish and coral structures\"}]}", "prompt_file": "04_underwater_manta.json", "prompt_id": "04_underwater_manta", "raw_frames_file": "results/benchmark/moe/original/raw_frames/04_underwater_manta.npy", "raw_frames_sha256": "68c6be000f9405a55cd6625cef632beb4484dcf5d1ce3157448c4093c0340ee3", "raw_frames_size_bytes": 87459968, "sample_count": 479, "sample_interval_seconds": 0.25, "seconds_per_frame": 1.9128099604516149, "seed": 4204, "shift": 3.0, "torch_allocated_after_mb": 103498.359375, "torch_peak_allocated_mb": 112469.43212890625, "torch_peak_reserved_mb": 124524.0, "torch_reserved_after_mb": 124524.0, "variant": "original", "width": 832}
|
| 5 |
+
{"batch_cfg": false, "cold_generation": false, "ffprobe": {"programs": [], "streams": [{"codec_name": "h264", "duration": "3.041667", "height": 480, "nb_frames": "73", "pix_fmt": "yuv420p", "r_frame_rate": "24/1", "width": 832}]}, "fps": 24, "generation_seconds": 139.8160271209199, "guidance_scale": 3.0, "height": 480, "model_key": "moe", "negative_prompt": "{\"universal_negative\": {\"visual_quality\": [\"low quality\", \"worst quality\", \"blurry\", \"pixelated\", \"jpeg artifacts\", \"low resolution\", \"unstable color\", \"color flicker\", \"underexposed\", \"overexposed\", \"invisible subject\", \"subject hidden in darkness\"], \"artistic_style\": [\"painting\", \"illustration\", \"drawing\", \"cartoon\", \"3d render\", \"cgi\", \"sketch\", \"digital art\"], \"composition_and_content\": [\"text\", \"watermark\", \"signature\", \"logo\", \"subtitles\", \"pillarboxed\", \"side bars\", \"portrait image in landscape frame\"], \"temporal_and_motion_stability\": [\"flickering\", \"jittery\", \"motion blur\", \"temporal inconsistency\", \"warping\", \"morphing\", \"incoherent motion\", \"unnatural movement\", \"static object with sudden jump\", \"frame-to-frame inconsistency\"], \"material_and_structure\": [\"plastic-like glass\", \"unrealistic texture\", \"deformed bottle\", \"liquid freezing improperly\", \"distorted reflections\"]}}", "null_cond_clone_zero": false, "num_frames": 73, "num_inference_steps": 40, "nvidia_smi_after": {"memory.total": "183359", "memory.used": "125646", "name": "NVIDIA B200", "power.draw": "260.91", "temperature.gpu": "43", "timestamp": "2026/07/10 08:39:38.091", "utilization.gpu": "0"}, "output_file": "results/benchmark/moe/original/videos/05_factory_battery_cells.mp4", "output_sha256": "8a08377ecb3877e3ab261ca1840450dfa28a18ccc6cc59679002da1ff7750b49", "output_size_bytes": 324116, "peak_disk_used_bytes": 188377776128.0, "peak_gpu_memory_used_mb": 125646.0, "peak_gpu_power_watts": 968.58, "peak_gpu_utilization_percent": 100.0, "peak_process_rss_bytes": 3052527616.0, "peak_system_used_bytes": 92903628800.0, "prompt": "{\"comprehensive_description\":{\"scene_content_description\":\"A clean high-tech factory line assembling transparent solid-state battery cells. Two white robotic arms move with precise synchronized motion over a brushed steel conveyor. One arm lowers a translucent rectangular cell into a copper test fixture while the other arm scans it with a blue light bar. Tiny reflections slide across glass safety panels, and status lights pulse softly. The scene should look like realistic industrial automation footage, not CGI, with crisp metal surfaces and controlled motion.\",\"camera_movement_description\":\"The camera is locked off in a medium-wide three-quarter view with a subtle mechanical vibration, emphasizing repeatable robotic motion.\"},\"camera_info\":{\"color\":\"White, brushed steel, copper, blue inspection light\",\"frame_size\":\"Medium wide\",\"shot_type_angle\":\"Three-quarter eye-level industrial view\",\"lens_size\":\"Medium lens\",\"composition\":\"Two robotic arms framing the battery cell at center\",\"lighting\":\"Even overhead factory lighting with blue scanner highlight\",\"lighting_type\":\"Artificial industrial lighting\"},\"world_knowledge\":[],\"prominent_elements\":[{\"name\":\"left robotic arm\",\"description\":\"A white six-axis robotic arm with a vacuum gripper holding a transparent rectangular battery cell.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.7s]\",\"action\":\"moves downward from upper left with the transparent cell held steady\"},{\"timestamp\":\"[0.7s - 1.4s]\",\"action\":\"places the cell into a copper test fixture\"},{\"timestamp\":\"[1.4s - 2.0s]\",\"action\":\"releases the cell and retracts slightly upward\"}],\"location\":\"left side moving toward center\",\"relative_size\":\"large\",\"shape_and_color\":\"white articulated segments with black joints\",\"texture\":\"smooth painted metal and rubber vacuum cups\",\"appearance_details\":\"visible cable routing, small green status LED\",\"relationship\":\"positions the battery cell for testing\",\"orientation\":\"angled downward toward the center fixture\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"right robotic scanner arm\",\"description\":\"A second white robotic arm carrying a rectangular blue inspection light bar.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.9s]\",\"action\":\"waits above the copper fixture with blue scanner light dim\"},{\"timestamp\":\"[0.9s - 1.6s]\",\"action\":\"sweeps the blue light bar across the transparent cell\"},{\"timestamp\":\"[1.6s - 2.0s]\",\"action\":\"pauses as the scanner light pulses once\"}],\"location\":\"right side above the conveyor\",\"relative_size\":\"large\",\"shape_and_color\":\"white arm with glowing blue rectangular scanner\",\"texture\":\"smooth metal casing and glass scanner cover\",\"appearance_details\":\"blue light strip, black joints, compact sensor module\",\"relationship\":\"inspects the cell after placement\",\"orientation\":\"angled left toward the battery cell\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"transparent battery cell and copper fixture\",\"description\":\"A clear rectangular solid-state cell seated into a copper test fixture on a steel conveyor.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.7s]\",\"action\":\"moves with the left arm above the fixture\"},{\"timestamp\":\"[0.7s - 1.4s]\",\"action\":\"is lowered precisely into the fixture\"},{\"timestamp\":\"[1.4s - 2.0s]\",\"action\":\"remains stationary while blue inspection light passes over it\"}],\"location\":\"center of the frame on the conveyor\",\"relative_size\":\"medium\",\"shape_and_color\":\"transparent rectangle with faint internal layers, copper fixture\",\"texture\":\"clear glass-like cell, brushed copper, polished steel\",\"appearance_details\":\"thin internal laminate layers, small alignment pins, clean conveyor surface\",\"relationship\":\"object being assembled and inspected\",\"orientation\":\"horizontal in the fixture\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"}]}", "prompt_file": "05_factory_battery_cells.json", "prompt_id": "05_factory_battery_cells", "raw_frames_file": "results/benchmark/moe/original/raw_frames/05_factory_battery_cells.npy", "raw_frames_sha256": "6107a35b979814f559baa8670e63c84f02b13c6f77c111203adb8ea199766c61", "raw_frames_size_bytes": 87459968, "sample_count": 473, "sample_interval_seconds": 0.25, "seconds_per_frame": 1.9152880427523273, "seed": 4205, "shift": 3.0, "torch_allocated_after_mb": 103497.9833984375, "torch_peak_allocated_mb": 112470.375, "torch_peak_reserved_mb": 124524.0, "torch_reserved_after_mb": 124524.0, "variant": "original", "width": 832}
|
benchmark/base/original/summary.json
ADDED
|
@@ -0,0 +1,370 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"cold_generation_seconds": 140.9023737310199,
|
| 3 |
+
"environment": {
|
| 4 |
+
"cuda": "12.8",
|
| 5 |
+
"gpu": "NVIDIA B200",
|
| 6 |
+
"lingbot_moe_expert_backend": "sglang_triton",
|
| 7 |
+
"python": "3.12.3 (main, Aug 14 2025, 17:47:21) [GCC 13.3.0]",
|
| 8 |
+
"torch": "2.8.0+cu128"
|
| 9 |
+
},
|
| 10 |
+
"generation_mean_seconds": 139.886719101551,
|
| 11 |
+
"hot_generation_mean_seconds": 139.63280544418376,
|
| 12 |
+
"lingbot_video_code_revision": "a2bb04b78edd848500dc27a26e035a95442ae186",
|
| 13 |
+
"load": {
|
| 14 |
+
"nvidia_smi_after": {
|
| 15 |
+
"memory.total": "183359",
|
| 16 |
+
"memory.used": "67234",
|
| 17 |
+
"name": "NVIDIA B200",
|
| 18 |
+
"power.draw": "249.62",
|
| 19 |
+
"temperature.gpu": "31",
|
| 20 |
+
"timestamp": "2026/07/10 08:27:42.682",
|
| 21 |
+
"utilization.gpu": "14"
|
| 22 |
+
},
|
| 23 |
+
"peak_disk_used_bytes": 188018044928.0,
|
| 24 |
+
"peak_gpu_memory_used_mb": 66692.0,
|
| 25 |
+
"peak_gpu_power_watts": 249.71,
|
| 26 |
+
"peak_gpu_utilization_percent": 19.0,
|
| 27 |
+
"peak_process_rss_bytes": 10081595392.0,
|
| 28 |
+
"peak_system_used_bytes": 78059479040.0,
|
| 29 |
+
"sample_count": 154,
|
| 30 |
+
"sample_interval_seconds": 0.25,
|
| 31 |
+
"seconds": 50.28812898299657,
|
| 32 |
+
"torch_allocated_after_mb": 66603.46142578125,
|
| 33 |
+
"torch_peak_allocated_mb": 66603.46142578125,
|
| 34 |
+
"torch_peak_reserved_mb": 66608.0,
|
| 35 |
+
"torch_reserved_after_mb": 66608.0
|
| 36 |
+
},
|
| 37 |
+
"model": "models/moe",
|
| 38 |
+
"model_key": "moe",
|
| 39 |
+
"peak_gpu_memory_used_mb": 126274.0,
|
| 40 |
+
"peak_process_rss_bytes": 3218972672.0,
|
| 41 |
+
"peak_system_used_bytes": 92903628800.0,
|
| 42 |
+
"peak_torch_allocated_mb": 112470.375,
|
| 43 |
+
"rows": [
|
| 44 |
+
{
|
| 45 |
+
"batch_cfg": false,
|
| 46 |
+
"cold_generation": true,
|
| 47 |
+
"ffprobe": {
|
| 48 |
+
"programs": [],
|
| 49 |
+
"streams": [
|
| 50 |
+
{
|
| 51 |
+
"codec_name": "h264",
|
| 52 |
+
"duration": "3.041667",
|
| 53 |
+
"height": 480,
|
| 54 |
+
"nb_frames": "73",
|
| 55 |
+
"pix_fmt": "yuv420p",
|
| 56 |
+
"r_frame_rate": "24/1",
|
| 57 |
+
"width": 832
|
| 58 |
+
}
|
| 59 |
+
]
|
| 60 |
+
},
|
| 61 |
+
"fps": 24,
|
| 62 |
+
"generation_seconds": 140.9023737310199,
|
| 63 |
+
"guidance_scale": 3.0,
|
| 64 |
+
"height": 480,
|
| 65 |
+
"model_key": "moe",
|
| 66 |
+
"negative_prompt": "{\"universal_negative\": {\"visual_quality\": [\"low quality\", \"worst quality\", \"blurry\", \"pixelated\", \"jpeg artifacts\", \"low resolution\", \"unstable color\", \"color flicker\", \"underexposed\", \"overexposed\", \"invisible subject\", \"subject hidden in darkness\"], \"artistic_style\": [\"painting\", \"illustration\", \"drawing\", \"cartoon\", \"3d render\", \"cgi\", \"sketch\", \"digital art\"], \"composition_and_content\": [\"text\", \"watermark\", \"signature\", \"logo\", \"subtitles\", \"pillarboxed\", \"side bars\", \"portrait image in landscape frame\"], \"temporal_and_motion_stability\": [\"flickering\", \"jittery\", \"motion blur\", \"temporal inconsistency\", \"warping\", \"morphing\", \"incoherent motion\", \"unnatural movement\", \"static object with sudden jump\", \"frame-to-frame inconsistency\"], \"material_and_structure\": [\"plastic-like glass\", \"unrealistic texture\", \"deformed bottle\", \"liquid freezing improperly\", \"distorted reflections\"]}}",
|
| 67 |
+
"null_cond_clone_zero": false,
|
| 68 |
+
"num_frames": 73,
|
| 69 |
+
"num_inference_steps": 40,
|
| 70 |
+
"nvidia_smi_after": {
|
| 71 |
+
"memory.total": "183359",
|
| 72 |
+
"memory.used": "125646",
|
| 73 |
+
"name": "NVIDIA B200",
|
| 74 |
+
"power.draw": "260.66",
|
| 75 |
+
"temperature.gpu": "43",
|
| 76 |
+
"timestamp": "2026/07/10 08:30:06.448",
|
| 77 |
+
"utilization.gpu": "0"
|
| 78 |
+
},
|
| 79 |
+
"output_file": "results/benchmark/moe/original/videos/01_arctic_research_drone.mp4",
|
| 80 |
+
"output_sha256": "7a548d09dd53755c1fdd7e759aae7cf92fd57d555bdc13f28023ede4ffe5f95e",
|
| 81 |
+
"output_size_bytes": 366044,
|
| 82 |
+
"peak_disk_used_bytes": 188024430592.0,
|
| 83 |
+
"peak_gpu_memory_used_mb": 125646.0,
|
| 84 |
+
"peak_gpu_power_watts": 977.6,
|
| 85 |
+
"peak_gpu_utilization_percent": 100.0,
|
| 86 |
+
"peak_process_rss_bytes": 2700222464.0,
|
| 87 |
+
"peak_system_used_bytes": 80392216576.0,
|
| 88 |
+
"prompt": "{\"comprehensive_description\":{\"scene_content_description\":\"A cinematic dusk shot of an Arctic research outpost on a wind-polished ice shelf under green aurora ribbons. A compact orange rover with a roof-mounted lidar mast drives slowly between blue instrument crates and slender weather antennas. Fine snow dust trails behind the tires, and warm light glows from the station windows against the cold cyan landscape. The atmosphere is realistic, high-detail documentary science footage with crisp snow texture, vapor in the air, and physically plausible shadows.\",\"camera_movement_description\":\"The camera begins as a low, smooth drone tracking shot behind the rover, then arcs gently left to reveal the glowing station and aurora-filled sky while keeping the rover dominant in frame.\"},\"camera_info\":{\"color\":\"Cold cyan with warm orange accents\",\"frame_size\":\"Wide\",\"shot_type_angle\":\"Low angle tracking shot\",\"lens_size\":\"Wide lens\",\"composition\":\"Rover in lower center, research station in the background, aurora above\",\"lighting\":\"Soft twilight with window glow\",\"lighting_type\":\"Natural dusk and practical interior lights\"},\"world_knowledge\":[],\"prominent_elements\":[{\"name\":\"orange autonomous research rover\",\"description\":\"A rugged six-wheeled scientific rover with orange body panels, black tires, a lidar mast, and small blinking status lights.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.7s]\",\"action\":\"drives forward slowly across packed snow, tires compressing the surface\"},{\"timestamp\":\"[0.7s - 1.5s]\",\"action\":\"turns slightly left as the lidar mast rotates\"},{\"timestamp\":\"[1.5s - 2.0s]\",\"action\":\"continues toward the lit station while fine snow trails behind\"}],\"location\":\"lower center of the frame\",\"relative_size\":\"dominant\",\"shape_and_color\":\"low rectangular orange body with black wheels and grey sensor mast\",\"texture\":\"matte painted metal, rubber tires, frost on edges\",\"appearance_details\":\"roof lidar, small antennas, narrow headlights, compact cargo rack\",\"relationship\":\"the main moving subject, traveling toward the research outpost\",\"orientation\":\"moving away from the camera and slightly left\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"Arctic research station\",\"description\":\"A modular polar station made of connected white containers with warm yellow light in the windows.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"remains stationary as wind-blown snow passes in front of it\"}],\"location\":\"middle background\",\"relative_size\":\"large\",\"shape_and_color\":\"rectangular white modules with dark seams and glowing windows\",\"texture\":\"frosted metal panels and glass\",\"appearance_details\":\"small stairs, railings, antennas, blue equipment crates nearby\",\"relationship\":\"destination of the rover\",\"orientation\":\"angled three-quarter view\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"aurora ribbons\",\"description\":\"Green aurora curtains stretching across the darkening polar sky.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"shimmer slowly and drift laterally across the sky\"}],\"location\":\"upper half of the frame\",\"relative_size\":\"large\",\"shape_and_color\":\"soft green luminous ribbons\",\"texture\":\"translucent atmospheric glow\",\"appearance_details\":\"layered folds and faint stars behind\",\"relationship\":\"dominates the sky above the station\",\"orientation\":\"horizontal sweeping arcs\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":true,\"number_of_objects\":\"many luminous bands\"}]}",
|
| 89 |
+
"prompt_file": "01_arctic_research_drone.json",
|
| 90 |
+
"prompt_id": "01_arctic_research_drone",
|
| 91 |
+
"raw_frames_file": "results/benchmark/moe/original/raw_frames/01_arctic_research_drone.npy",
|
| 92 |
+
"raw_frames_sha256": "1c1cad878f2c8f39889f0acb71d6ebaf1b12222ff43ddcd8419e54e9cf00d590",
|
| 93 |
+
"raw_frames_size_bytes": 87459968,
|
| 94 |
+
"sample_count": 488,
|
| 95 |
+
"sample_interval_seconds": 0.25,
|
| 96 |
+
"seconds_per_frame": 1.9301695031646562,
|
| 97 |
+
"seed": 4201,
|
| 98 |
+
"shift": 3.0,
|
| 99 |
+
"torch_allocated_after_mb": 103497.9833984375,
|
| 100 |
+
"torch_peak_allocated_mb": 112469.38330078125,
|
| 101 |
+
"torch_peak_reserved_mb": 124524.0,
|
| 102 |
+
"torch_reserved_after_mb": 124524.0,
|
| 103 |
+
"variant": "original",
|
| 104 |
+
"width": 832
|
| 105 |
+
},
|
| 106 |
+
{
|
| 107 |
+
"batch_cfg": false,
|
| 108 |
+
"cold_generation": false,
|
| 109 |
+
"ffprobe": {
|
| 110 |
+
"programs": [],
|
| 111 |
+
"streams": [
|
| 112 |
+
{
|
| 113 |
+
"codec_name": "h264",
|
| 114 |
+
"duration": "3.041667",
|
| 115 |
+
"height": 480,
|
| 116 |
+
"nb_frames": "73",
|
| 117 |
+
"pix_fmt": "yuv420p",
|
| 118 |
+
"r_frame_rate": "24/1",
|
| 119 |
+
"width": 832
|
| 120 |
+
}
|
| 121 |
+
]
|
| 122 |
+
},
|
| 123 |
+
"fps": 24,
|
| 124 |
+
"generation_seconds": 139.39877224492375,
|
| 125 |
+
"guidance_scale": 3.0,
|
| 126 |
+
"height": 480,
|
| 127 |
+
"model_key": "moe",
|
| 128 |
+
"negative_prompt": "{\"universal_negative\": {\"visual_quality\": [\"low quality\", \"worst quality\", \"blurry\", \"pixelated\", \"jpeg artifacts\", \"low resolution\", \"unstable color\", \"color flicker\", \"underexposed\", \"overexposed\", \"invisible subject\", \"subject hidden in darkness\"], \"artistic_style\": [\"painting\", \"illustration\", \"drawing\", \"cartoon\", \"3d render\", \"cgi\", \"sketch\", \"digital art\"], \"composition_and_content\": [\"text\", \"watermark\", \"signature\", \"logo\", \"subtitles\", \"pillarboxed\", \"side bars\", \"portrait image in landscape frame\"], \"temporal_and_motion_stability\": [\"flickering\", \"jittery\", \"motion blur\", \"temporal inconsistency\", \"warping\", \"morphing\", \"incoherent motion\", \"unnatural movement\", \"static object with sudden jump\", \"frame-to-frame inconsistency\"], \"material_and_structure\": [\"plastic-like glass\", \"unrealistic texture\", \"deformed bottle\", \"liquid freezing improperly\", \"distorted reflections\"]}}",
|
| 129 |
+
"null_cond_clone_zero": false,
|
| 130 |
+
"num_frames": 73,
|
| 131 |
+
"num_inference_steps": 40,
|
| 132 |
+
"nvidia_smi_after": {
|
| 133 |
+
"memory.total": "183359",
|
| 134 |
+
"memory.used": "125646",
|
| 135 |
+
"name": "NVIDIA B200",
|
| 136 |
+
"power.draw": "260.70",
|
| 137 |
+
"temperature.gpu": "43",
|
| 138 |
+
"timestamp": "2026/07/10 08:32:29.979",
|
| 139 |
+
"utilization.gpu": "0"
|
| 140 |
+
},
|
| 141 |
+
"output_file": "results/benchmark/moe/original/videos/02_macro_pastry_glaze.mp4",
|
| 142 |
+
"output_sha256": "0073c4f94ed8fc0b9bfaa93966ac627b8a66f31efe9547a78bc6934ee2ecfd82",
|
| 143 |
+
"output_size_bytes": 396167,
|
| 144 |
+
"peak_disk_used_bytes": 188112715776.0,
|
| 145 |
+
"peak_gpu_memory_used_mb": 126274.0,
|
| 146 |
+
"peak_gpu_power_watts": 969.78,
|
| 147 |
+
"peak_gpu_utilization_percent": 100.0,
|
| 148 |
+
"peak_process_rss_bytes": 3164041216.0,
|
| 149 |
+
"peak_system_used_bytes": 90717732864.0,
|
| 150 |
+
"prompt": "{\"comprehensive_description\":{\"scene_content_description\":\"An extreme macro food cinematography shot in a professional pastry kitchen. A glossy dark chocolate dome dessert sits on a rotating metal turntable. A thin stream of amber caramel glaze pours from above, flowing over the curved surface in slow ribbons and revealing tiny reflections from softbox lights. Crushed pistachio dust and edible gold flakes sit around the base on a matte black plate. The look is photorealistic, tactile, high-speed macro video with shallow depth of field and no text.\",\"camera_movement_description\":\"The camera holds a close macro angle with a subtle push-in while the dessert rotates clockwise on the turntable.\"},\"camera_info\":{\"color\":\"Deep brown, amber, green pistachio, gold highlights\",\"frame_size\":\"Extreme close up\",\"shot_type_angle\":\"Slight high angle\",\"lens_size\":\"Macro lens\",\"composition\":\"Dessert centered with glaze entering from top\",\"lighting\":\"Large softbox reflections\",\"lighting_type\":\"Controlled studio kitchen lighting\"},\"world_knowledge\":[],\"prominent_elements\":[{\"name\":\"chocolate dome dessert\",\"description\":\"A small mirror-glazed chocolate dome on a black plate, surrounded by pistachio dust and gold flakes.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"rotates slowly clockwise on a metal turntable\"}],\"location\":\"center of the frame\",\"relative_size\":\"dominant\",\"shape_and_color\":\"hemispherical dark chocolate dome on round black plate\",\"texture\":\"mirror-gloss surface with smooth curved reflections\",\"appearance_details\":\"tiny gold flakes, green pistachio crumbs, clean pastry plating\",\"relationship\":\"receives the caramel glaze\",\"orientation\":\"upright on the turntable\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"caramel glaze stream\",\"description\":\"A viscous amber caramel stream pouring from above and spreading over the dessert.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.8s]\",\"action\":\"falls as a narrow glossy stream onto the crown of the dome\"},{\"timestamp\":\"[0.8s - 1.6s]\",\"action\":\"forms slow ribbons that slide down the curved chocolate surface\"},{\"timestamp\":\"[1.6s - 2.0s]\",\"action\":\"collects at the base in a thin shiny ring\"}],\"location\":\"entering from top center and flowing over the dessert\",\"relative_size\":\"medium\",\"shape_and_color\":\"thin amber liquid ribbons\",\"texture\":\"viscous, glossy, translucent caramel\",\"appearance_details\":\"bright highlights and tiny bubbles\",\"relationship\":\"coats the chocolate dome\",\"orientation\":\"vertical stream becoming downward trails\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"}]}",
|
| 151 |
+
"prompt_file": "02_macro_pastry_glaze.json",
|
| 152 |
+
"prompt_id": "02_macro_pastry_glaze",
|
| 153 |
+
"raw_frames_file": "results/benchmark/moe/original/raw_frames/02_macro_pastry_glaze.npy",
|
| 154 |
+
"raw_frames_sha256": "c96f6b8795b970785ca2c50c383ab9e6055b18c05496bbe0873aa67dbea063ec",
|
| 155 |
+
"raw_frames_size_bytes": 87459968,
|
| 156 |
+
"sample_count": 463,
|
| 157 |
+
"sample_interval_seconds": 0.25,
|
| 158 |
+
"seconds_per_frame": 1.9095722225332021,
|
| 159 |
+
"seed": 4202,
|
| 160 |
+
"shift": 3.0,
|
| 161 |
+
"torch_allocated_after_mb": 103497.9833984375,
|
| 162 |
+
"torch_peak_allocated_mb": 112468.59130859375,
|
| 163 |
+
"torch_peak_reserved_mb": 124524.0,
|
| 164 |
+
"torch_reserved_after_mb": 124524.0,
|
| 165 |
+
"variant": "original",
|
| 166 |
+
"width": 832
|
| 167 |
+
},
|
| 168 |
+
{
|
| 169 |
+
"batch_cfg": false,
|
| 170 |
+
"cold_generation": false,
|
| 171 |
+
"ffprobe": {
|
| 172 |
+
"programs": [],
|
| 173 |
+
"streams": [
|
| 174 |
+
{
|
| 175 |
+
"codec_name": "h264",
|
| 176 |
+
"duration": "3.041667",
|
| 177 |
+
"height": 480,
|
| 178 |
+
"nb_frames": "73",
|
| 179 |
+
"pix_fmt": "yuv420p",
|
| 180 |
+
"r_frame_rate": "24/1",
|
| 181 |
+
"width": 832
|
| 182 |
+
}
|
| 183 |
+
]
|
| 184 |
+
},
|
| 185 |
+
"fps": 24,
|
| 186 |
+
"generation_seconds": 139.6812952979235,
|
| 187 |
+
"guidance_scale": 3.0,
|
| 188 |
+
"height": 480,
|
| 189 |
+
"model_key": "moe",
|
| 190 |
+
"negative_prompt": "{\"universal_negative\": {\"visual_quality\": [\"low quality\", \"worst quality\", \"blurry\", \"pixelated\", \"jpeg artifacts\", \"low resolution\", \"unstable color\", \"color flicker\", \"underexposed\", \"overexposed\", \"invisible subject\", \"subject hidden in darkness\"], \"artistic_style\": [\"painting\", \"illustration\", \"drawing\", \"cartoon\", \"3d render\", \"cgi\", \"sketch\", \"digital art\"], \"composition_and_content\": [\"text\", \"watermark\", \"signature\", \"logo\", \"subtitles\", \"pillarboxed\", \"side bars\", \"portrait image in landscape frame\"], \"temporal_and_motion_stability\": [\"flickering\", \"jittery\", \"motion blur\", \"temporal inconsistency\", \"warping\", \"morphing\", \"incoherent motion\", \"unnatural movement\", \"static object with sudden jump\", \"frame-to-frame inconsistency\"], \"material_and_structure\": [\"plastic-like glass\", \"unrealistic texture\", \"deformed bottle\", \"liquid freezing improperly\", \"distorted reflections\"]}}",
|
| 191 |
+
"null_cond_clone_zero": false,
|
| 192 |
+
"num_frames": 73,
|
| 193 |
+
"num_inference_steps": 40,
|
| 194 |
+
"nvidia_smi_after": {
|
| 195 |
+
"memory.total": "183359",
|
| 196 |
+
"memory.used": "125646",
|
| 197 |
+
"name": "NVIDIA B200",
|
| 198 |
+
"power.draw": "260.53",
|
| 199 |
+
"temperature.gpu": "43",
|
| 200 |
+
"timestamp": "2026/07/10 08:34:52.790",
|
| 201 |
+
"utilization.gpu": "0"
|
| 202 |
+
},
|
| 203 |
+
"output_file": "results/benchmark/moe/original/videos/03_rain_market_lanterns.mp4",
|
| 204 |
+
"output_sha256": "753f2016262fbd6a3644b28ad4f6243e15e8b55ef9474616a9c7d636f474090b",
|
| 205 |
+
"output_size_bytes": 430040,
|
| 206 |
+
"peak_disk_used_bytes": 188200804352.0,
|
| 207 |
+
"peak_gpu_memory_used_mb": 125646.0,
|
| 208 |
+
"peak_gpu_power_watts": 972.57,
|
| 209 |
+
"peak_gpu_utilization_percent": 100.0,
|
| 210 |
+
"peak_process_rss_bytes": 3218972672.0,
|
| 211 |
+
"peak_system_used_bytes": 91209596928.0,
|
| 212 |
+
"prompt": "{\"comprehensive_description\":{\"scene_content_description\":\"A lively rainy night market in a narrow old-city street, filmed like naturalistic cinema. Red paper lanterns hang from wooden stalls, steam rises from bowls of soup, and wet cobblestones mirror orange and teal reflections. A vendor in a dark apron slides a steaming bowl across a counter to a traveler in a blue raincoat. Umbrellas pass in the foreground, partially occluding the scene for a moment. The mood is warm, humid, detailed, and grounded, with no readable signs or subtitles.\",\"camera_movement_description\":\"The camera performs a slow lateral dolly from left to right at counter height, with foreground umbrellas briefly crossing the lens.\"},\"camera_info\":{\"color\":\"Warm lantern orange, teal rain reflections, red accents\",\"frame_size\":\"Medium wide\",\"shot_type_angle\":\"Eye level\",\"lens_size\":\"Medium lens\",\"composition\":\"Vendor stall centered, traveler on right, passing umbrellas in foreground\",\"lighting\":\"Lantern light mixed with rain reflections\",\"lighting_type\":\"Night practical lighting\"},\"world_knowledge\":[],\"prominent_elements\":[{\"name\":\"food vendor\",\"description\":\"A middle-aged street vendor wearing a dark apron and working behind a steaming wooden food stall.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.8s]\",\"action\":\"lifts a white ceramic bowl from a steaming pot\"},{\"timestamp\":\"[0.8s - 1.6s]\",\"action\":\"slides the bowl carefully across the counter\"},{\"timestamp\":\"[1.6s - 2.0s]\",\"action\":\"wipes condensation from the counter with a cloth\"}],\"location\":\"center-left behind the stall counter\",\"relative_size\":\"large\",\"shape_and_color\":\"human figure in dark apron under warm lantern light\",\"texture\":\"wet fabric, skin highlights, wood grain around the stall\",\"appearance_details\":\"rolled sleeves, focused expression, steam around hands\",\"relationship\":\"serves food to the traveler\",\"orientation\":\"facing right toward the customer\",\"pose\":\"leaning slightly forward over the counter\",\"expression\":\"focused and calm\",\"clothing\":\"dark apron over a muted shirt\",\"gender\":\"\",\"skin_tone_and_texture\":\"natural skin with rain-lit highlights\"},{\"name\":\"traveler in blue raincoat\",\"description\":\"A traveler wearing a blue hooded raincoat, waiting at the stall with wet sleeves and a folded umbrella.\",\"actions\":[{\"timestamp\":\"[0.0s - 1.2s]\",\"action\":\"waits with hands near the counter, watching the bowl\"},{\"timestamp\":\"[1.2s - 2.0s]\",\"action\":\"reaches forward to receive the steaming bowl\"}],\"location\":\"right side of the frame\",\"relative_size\":\"medium\",\"shape_and_color\":\"blue hooded raincoat silhouette\",\"texture\":\"water beads on waterproof fabric\",\"appearance_details\":\"hood up, folded umbrella tucked under one arm\",\"relationship\":\"customer receiving food from the vendor\",\"orientation\":\"facing left toward the stall\",\"pose\":\"standing close to the counter\",\"expression\":\"expectant and tired\",\"clothing\":\"blue raincoat\",\"gender\":\"\",\"skin_tone_and_texture\":\"\"},{\"name\":\"rain and lantern reflections\",\"description\":\"Falling rain, steam, and glowing reflections on wet cobblestones.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"rain falls continuously while reflections ripple on the street\"}],\"location\":\"foreground and background throughout the frame\",\"relative_size\":\"large\",\"shape_and_color\":\"thin rain streaks, orange and teal reflected pools\",\"texture\":\"wet stone, vapor, glossy water\",\"appearance_details\":\"soft steam plumes, umbrellas passing close to lens\",\"relationship\":\"sets the atmosphere around the market\",\"orientation\":\"vertical rainfall and horizontal street reflections\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":true,\"number_of_objects\":\"many raindrops and reflections\"}]}",
|
| 213 |
+
"prompt_file": "03_rain_market_lanterns.json",
|
| 214 |
+
"prompt_id": "03_rain_market_lanterns",
|
| 215 |
+
"raw_frames_file": "results/benchmark/moe/original/raw_frames/03_rain_market_lanterns.npy",
|
| 216 |
+
"raw_frames_sha256": "8b0391e7276b04d2d802ce81b1f1a32d5e4cd0c976f55d2466daaa3a6c838a63",
|
| 217 |
+
"raw_frames_size_bytes": 87459968,
|
| 218 |
+
"sample_count": 486,
|
| 219 |
+
"sample_interval_seconds": 0.25,
|
| 220 |
+
"seconds_per_frame": 1.9134424013414177,
|
| 221 |
+
"seed": 4203,
|
| 222 |
+
"shift": 3.0,
|
| 223 |
+
"torch_allocated_after_mb": 103497.9833984375,
|
| 224 |
+
"torch_peak_allocated_mb": 112469.681640625,
|
| 225 |
+
"torch_peak_reserved_mb": 124524.0,
|
| 226 |
+
"torch_reserved_after_mb": 124524.0,
|
| 227 |
+
"variant": "original",
|
| 228 |
+
"width": 832
|
| 229 |
+
},
|
| 230 |
+
{
|
| 231 |
+
"batch_cfg": false,
|
| 232 |
+
"cold_generation": false,
|
| 233 |
+
"ffprobe": {
|
| 234 |
+
"programs": [],
|
| 235 |
+
"streams": [
|
| 236 |
+
{
|
| 237 |
+
"codec_name": "h264",
|
| 238 |
+
"duration": "3.041667",
|
| 239 |
+
"height": 480,
|
| 240 |
+
"nb_frames": "73",
|
| 241 |
+
"pix_fmt": "yuv420p",
|
| 242 |
+
"r_frame_rate": "24/1",
|
| 243 |
+
"width": 832
|
| 244 |
+
}
|
| 245 |
+
]
|
| 246 |
+
},
|
| 247 |
+
"fps": 24,
|
| 248 |
+
"generation_seconds": 139.6351271129679,
|
| 249 |
+
"guidance_scale": 3.0,
|
| 250 |
+
"height": 480,
|
| 251 |
+
"model_key": "moe",
|
| 252 |
+
"negative_prompt": "{\"universal_negative\": {\"visual_quality\": [\"low quality\", \"worst quality\", \"blurry\", \"pixelated\", \"jpeg artifacts\", \"low resolution\", \"unstable color\", \"color flicker\", \"underexposed\", \"overexposed\", \"invisible subject\", \"subject hidden in darkness\"], \"artistic_style\": [\"painting\", \"illustration\", \"drawing\", \"cartoon\", \"3d render\", \"cgi\", \"sketch\", \"digital art\"], \"composition_and_content\": [\"text\", \"watermark\", \"signature\", \"logo\", \"subtitles\", \"pillarboxed\", \"side bars\", \"portrait image in landscape frame\"], \"temporal_and_motion_stability\": [\"flickering\", \"jittery\", \"motion blur\", \"temporal inconsistency\", \"warping\", \"morphing\", \"incoherent motion\", \"unnatural movement\", \"static object with sudden jump\", \"frame-to-frame inconsistency\"], \"material_and_structure\": [\"plastic-like glass\", \"unrealistic texture\", \"deformed bottle\", \"liquid freezing improperly\", \"distorted reflections\"]}}",
|
| 253 |
+
"null_cond_clone_zero": false,
|
| 254 |
+
"num_frames": 73,
|
| 255 |
+
"num_inference_steps": 40,
|
| 256 |
+
"nvidia_smi_after": {
|
| 257 |
+
"memory.total": "183359",
|
| 258 |
+
"memory.used": "125646",
|
| 259 |
+
"name": "NVIDIA B200",
|
| 260 |
+
"power.draw": "261.37",
|
| 261 |
+
"temperature.gpu": "43",
|
| 262 |
+
"timestamp": "2026/07/10 08:37:15.363",
|
| 263 |
+
"utilization.gpu": "0"
|
| 264 |
+
},
|
| 265 |
+
"output_file": "results/benchmark/moe/original/videos/04_underwater_manta.mp4",
|
| 266 |
+
"output_sha256": "dcd3ffe1e7dd1f72f86e256d0db51acfb6661b880bfb0b2a5b798c9609cae24a",
|
| 267 |
+
"output_size_bytes": 359114,
|
| 268 |
+
"peak_disk_used_bytes": 188289716224.0,
|
| 269 |
+
"peak_gpu_memory_used_mb": 125646.0,
|
| 270 |
+
"peak_gpu_power_watts": 978.59,
|
| 271 |
+
"peak_gpu_utilization_percent": 100.0,
|
| 272 |
+
"peak_process_rss_bytes": 3070435328.0,
|
| 273 |
+
"peak_system_used_bytes": 91173777408.0,
|
| 274 |
+
"prompt": "{\"comprehensive_description\":{\"scene_content_description\":\"A tranquil underwater wildlife scene in clear tropical water. A large manta ray glides above a colorful coral garden while a scuba diver hovers several meters behind, holding a small camera rig. Sunbeams ripple through the surface, schools of tiny silver fish split and rejoin around the manta, and suspended particles drift slowly. The image should feel like realistic underwater documentary footage with natural motion, soft blue-green color, and no fantasy elements.\",\"camera_movement_description\":\"The camera tracks smoothly beside the manta ray from a slightly lower angle, drifting forward with the current.\"},\"camera_info\":{\"color\":\"Blue-green water with coral reds and yellows\",\"frame_size\":\"Wide\",\"shot_type_angle\":\"Slight low angle\",\"lens_size\":\"Wide underwater lens\",\"composition\":\"Manta ray crossing the upper center, coral below, diver in background\",\"lighting\":\"Dappled sunlight\",\"lighting_type\":\"Natural underwater daylight\"},\"world_knowledge\":[],\"prominent_elements\":[{\"name\":\"manta ray\",\"description\":\"A large manta ray with broad triangular fins, dark top surface, and pale underside.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.8s]\",\"action\":\"glides from left to right with slow wing-like fin movement\"},{\"timestamp\":\"[0.8s - 1.5s]\",\"action\":\"tilts slightly upward as sunbeams cross its back\"},{\"timestamp\":\"[1.5s - 2.0s]\",\"action\":\"continues forward while small fish scatter around it\"}],\"location\":\"upper center moving toward right\",\"relative_size\":\"dominant\",\"shape_and_color\":\"wide diamond-like silhouette, dark grey top, pale underside\",\"texture\":\"smooth skin with subtle mottling\",\"appearance_details\":\"cephalic fins near mouth, long tail trailing behind\",\"relationship\":\"main wildlife subject of the video\",\"orientation\":\"moving left to right and slightly upward\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"scuba diver\",\"description\":\"A diver in black wetsuit and fins, carrying a small underwater camera rig.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"hovers calmly in the background, exhaling small bubble streams\"}],\"location\":\"mid-background behind the manta ray\",\"relative_size\":\"small\",\"shape_and_color\":\"human figure in black wetsuit with silver air tank\",\"texture\":\"neoprene suit, metal tank, glass mask\",\"appearance_details\":\"fins, mask, regulator, compact camera lights\",\"relationship\":\"observes the manta ray without touching it\",\"orientation\":\"facing the manta ray\",\"pose\":\"horizontal hover\",\"expression\":\"\",\"clothing\":\"black scuba gear\",\"gender\":\"\",\"skin_tone_and_texture\":\"\"},{\"name\":\"coral reef and fish\",\"description\":\"A colorful coral garden with small silver fish schooling above it.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"fish swirl and split around the manta while coral remains fixed\"}],\"location\":\"lower half of the frame\",\"relative_size\":\"large\",\"shape_and_color\":\"branching corals in red, yellow, and muted purple; silver fish clusters\",\"texture\":\"rough coral, shimmering fish scales\",\"appearance_details\":\"sand patches, sea fans, drifting particles\",\"relationship\":\"environment below the manta ray\",\"orientation\":\"reef spreads horizontally across the bottom\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":true,\"number_of_objects\":\"many fish and coral structures\"}]}",
|
| 275 |
+
"prompt_file": "04_underwater_manta.json",
|
| 276 |
+
"prompt_id": "04_underwater_manta",
|
| 277 |
+
"raw_frames_file": "results/benchmark/moe/original/raw_frames/04_underwater_manta.npy",
|
| 278 |
+
"raw_frames_sha256": "68c6be000f9405a55cd6625cef632beb4484dcf5d1ce3157448c4093c0340ee3",
|
| 279 |
+
"raw_frames_size_bytes": 87459968,
|
| 280 |
+
"sample_count": 479,
|
| 281 |
+
"sample_interval_seconds": 0.25,
|
| 282 |
+
"seconds_per_frame": 1.9128099604516149,
|
| 283 |
+
"seed": 4204,
|
| 284 |
+
"shift": 3.0,
|
| 285 |
+
"torch_allocated_after_mb": 103498.359375,
|
| 286 |
+
"torch_peak_allocated_mb": 112469.43212890625,
|
| 287 |
+
"torch_peak_reserved_mb": 124524.0,
|
| 288 |
+
"torch_reserved_after_mb": 124524.0,
|
| 289 |
+
"variant": "original",
|
| 290 |
+
"width": 832
|
| 291 |
+
},
|
| 292 |
+
{
|
| 293 |
+
"batch_cfg": false,
|
| 294 |
+
"cold_generation": false,
|
| 295 |
+
"ffprobe": {
|
| 296 |
+
"programs": [],
|
| 297 |
+
"streams": [
|
| 298 |
+
{
|
| 299 |
+
"codec_name": "h264",
|
| 300 |
+
"duration": "3.041667",
|
| 301 |
+
"height": 480,
|
| 302 |
+
"nb_frames": "73",
|
| 303 |
+
"pix_fmt": "yuv420p",
|
| 304 |
+
"r_frame_rate": "24/1",
|
| 305 |
+
"width": 832
|
| 306 |
+
}
|
| 307 |
+
]
|
| 308 |
+
},
|
| 309 |
+
"fps": 24,
|
| 310 |
+
"generation_seconds": 139.8160271209199,
|
| 311 |
+
"guidance_scale": 3.0,
|
| 312 |
+
"height": 480,
|
| 313 |
+
"model_key": "moe",
|
| 314 |
+
"negative_prompt": "{\"universal_negative\": {\"visual_quality\": [\"low quality\", \"worst quality\", \"blurry\", \"pixelated\", \"jpeg artifacts\", \"low resolution\", \"unstable color\", \"color flicker\", \"underexposed\", \"overexposed\", \"invisible subject\", \"subject hidden in darkness\"], \"artistic_style\": [\"painting\", \"illustration\", \"drawing\", \"cartoon\", \"3d render\", \"cgi\", \"sketch\", \"digital art\"], \"composition_and_content\": [\"text\", \"watermark\", \"signature\", \"logo\", \"subtitles\", \"pillarboxed\", \"side bars\", \"portrait image in landscape frame\"], \"temporal_and_motion_stability\": [\"flickering\", \"jittery\", \"motion blur\", \"temporal inconsistency\", \"warping\", \"morphing\", \"incoherent motion\", \"unnatural movement\", \"static object with sudden jump\", \"frame-to-frame inconsistency\"], \"material_and_structure\": [\"plastic-like glass\", \"unrealistic texture\", \"deformed bottle\", \"liquid freezing improperly\", \"distorted reflections\"]}}",
|
| 315 |
+
"null_cond_clone_zero": false,
|
| 316 |
+
"num_frames": 73,
|
| 317 |
+
"num_inference_steps": 40,
|
| 318 |
+
"nvidia_smi_after": {
|
| 319 |
+
"memory.total": "183359",
|
| 320 |
+
"memory.used": "125646",
|
| 321 |
+
"name": "NVIDIA B200",
|
| 322 |
+
"power.draw": "260.91",
|
| 323 |
+
"temperature.gpu": "43",
|
| 324 |
+
"timestamp": "2026/07/10 08:39:38.091",
|
| 325 |
+
"utilization.gpu": "0"
|
| 326 |
+
},
|
| 327 |
+
"output_file": "results/benchmark/moe/original/videos/05_factory_battery_cells.mp4",
|
| 328 |
+
"output_sha256": "8a08377ecb3877e3ab261ca1840450dfa28a18ccc6cc59679002da1ff7750b49",
|
| 329 |
+
"output_size_bytes": 324116,
|
| 330 |
+
"peak_disk_used_bytes": 188377776128.0,
|
| 331 |
+
"peak_gpu_memory_used_mb": 125646.0,
|
| 332 |
+
"peak_gpu_power_watts": 968.58,
|
| 333 |
+
"peak_gpu_utilization_percent": 100.0,
|
| 334 |
+
"peak_process_rss_bytes": 3052527616.0,
|
| 335 |
+
"peak_system_used_bytes": 92903628800.0,
|
| 336 |
+
"prompt": "{\"comprehensive_description\":{\"scene_content_description\":\"A clean high-tech factory line assembling transparent solid-state battery cells. Two white robotic arms move with precise synchronized motion over a brushed steel conveyor. One arm lowers a translucent rectangular cell into a copper test fixture while the other arm scans it with a blue light bar. Tiny reflections slide across glass safety panels, and status lights pulse softly. The scene should look like realistic industrial automation footage, not CGI, with crisp metal surfaces and controlled motion.\",\"camera_movement_description\":\"The camera is locked off in a medium-wide three-quarter view with a subtle mechanical vibration, emphasizing repeatable robotic motion.\"},\"camera_info\":{\"color\":\"White, brushed steel, copper, blue inspection light\",\"frame_size\":\"Medium wide\",\"shot_type_angle\":\"Three-quarter eye-level industrial view\",\"lens_size\":\"Medium lens\",\"composition\":\"Two robotic arms framing the battery cell at center\",\"lighting\":\"Even overhead factory lighting with blue scanner highlight\",\"lighting_type\":\"Artificial industrial lighting\"},\"world_knowledge\":[],\"prominent_elements\":[{\"name\":\"left robotic arm\",\"description\":\"A white six-axis robotic arm with a vacuum gripper holding a transparent rectangular battery cell.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.7s]\",\"action\":\"moves downward from upper left with the transparent cell held steady\"},{\"timestamp\":\"[0.7s - 1.4s]\",\"action\":\"places the cell into a copper test fixture\"},{\"timestamp\":\"[1.4s - 2.0s]\",\"action\":\"releases the cell and retracts slightly upward\"}],\"location\":\"left side moving toward center\",\"relative_size\":\"large\",\"shape_and_color\":\"white articulated segments with black joints\",\"texture\":\"smooth painted metal and rubber vacuum cups\",\"appearance_details\":\"visible cable routing, small green status LED\",\"relationship\":\"positions the battery cell for testing\",\"orientation\":\"angled downward toward the center fixture\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"right robotic scanner arm\",\"description\":\"A second white robotic arm carrying a rectangular blue inspection light bar.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.9s]\",\"action\":\"waits above the copper fixture with blue scanner light dim\"},{\"timestamp\":\"[0.9s - 1.6s]\",\"action\":\"sweeps the blue light bar across the transparent cell\"},{\"timestamp\":\"[1.6s - 2.0s]\",\"action\":\"pauses as the scanner light pulses once\"}],\"location\":\"right side above the conveyor\",\"relative_size\":\"large\",\"shape_and_color\":\"white arm with glowing blue rectangular scanner\",\"texture\":\"smooth metal casing and glass scanner cover\",\"appearance_details\":\"blue light strip, black joints, compact sensor module\",\"relationship\":\"inspects the cell after placement\",\"orientation\":\"angled left toward the battery cell\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"transparent battery cell and copper fixture\",\"description\":\"A clear rectangular solid-state cell seated into a copper test fixture on a steel conveyor.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.7s]\",\"action\":\"moves with the left arm above the fixture\"},{\"timestamp\":\"[0.7s - 1.4s]\",\"action\":\"is lowered precisely into the fixture\"},{\"timestamp\":\"[1.4s - 2.0s]\",\"action\":\"remains stationary while blue inspection light passes over it\"}],\"location\":\"center of the frame on the conveyor\",\"relative_size\":\"medium\",\"shape_and_color\":\"transparent rectangle with faint internal layers, copper fixture\",\"texture\":\"clear glass-like cell, brushed copper, polished steel\",\"appearance_details\":\"thin internal laminate layers, small alignment pins, clean conveyor surface\",\"relationship\":\"object being assembled and inspected\",\"orientation\":\"horizontal in the fixture\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"}]}",
|
| 337 |
+
"prompt_file": "05_factory_battery_cells.json",
|
| 338 |
+
"prompt_id": "05_factory_battery_cells",
|
| 339 |
+
"raw_frames_file": "results/benchmark/moe/original/raw_frames/05_factory_battery_cells.npy",
|
| 340 |
+
"raw_frames_sha256": "6107a35b979814f559baa8670e63c84f02b13c6f77c111203adb8ea199766c61",
|
| 341 |
+
"raw_frames_size_bytes": 87459968,
|
| 342 |
+
"sample_count": 473,
|
| 343 |
+
"sample_interval_seconds": 0.25,
|
| 344 |
+
"seconds_per_frame": 1.9152880427523273,
|
| 345 |
+
"seed": 4205,
|
| 346 |
+
"shift": 3.0,
|
| 347 |
+
"torch_allocated_after_mb": 103497.9833984375,
|
| 348 |
+
"torch_peak_allocated_mb": 112470.375,
|
| 349 |
+
"torch_peak_reserved_mb": 124524.0,
|
| 350 |
+
"torch_reserved_after_mb": 124524.0,
|
| 351 |
+
"variant": "original",
|
| 352 |
+
"width": 832
|
| 353 |
+
}
|
| 354 |
+
],
|
| 355 |
+
"sdnq_revision": null,
|
| 356 |
+
"settings": {
|
| 357 |
+
"batch_cfg": false,
|
| 358 |
+
"fps": 24,
|
| 359 |
+
"guidance_scale": 3.0,
|
| 360 |
+
"height": 480,
|
| 361 |
+
"null_cond_clone_zero": false,
|
| 362 |
+
"num_frames": 73,
|
| 363 |
+
"num_inference_steps": 40,
|
| 364 |
+
"shift": 3.0,
|
| 365 |
+
"width": 832
|
| 366 |
+
},
|
| 367 |
+
"source_repo_id": "robbyant/lingbot-video-moe-30b-a3b",
|
| 368 |
+
"source_revision": "f2e538f64afe00cc4ae674db2aeb52e2945edfd5",
|
| 369 |
+
"variant": "original"
|
| 370 |
+
}
|
benchmark/base/sdnq/metrics.csv
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
model_key,variant,prompt_id,seed,generation_seconds,seconds_per_frame,torch_peak_allocated_mb,torch_peak_reserved_mb,peak_gpu_memory_used_mb,peak_gpu_utilization_percent,peak_process_rss_bytes,peak_system_used_bytes,output_size_bytes,output_sha256
|
| 2 |
+
moe,sdnq,01_arctic_research_drone,4201,159.77318301890045,2.188673739984938,36067.29833984375,40756.0,41878.0,100.0,2920222720.0,138935304192.0,339957,d040042bdb79aacdcafe80a2c4bd621271e3056823a84621c25ee8fd4fff594b
|
| 3 |
+
moe,sdnq,02_macro_pastry_glaze,4202,158.2769890979398,2.1681779328484905,36065.68798828125,40756.0,42504.0,100.0,3105587200.0,90820800512.0,367773,bcb2caabb9e29427019ef3dcc7790a89910c1b570e65f85bac3f8a95dc1912db
|
| 4 |
+
moe,sdnq,03_rain_market_lanterns,4203,158.72098272107542,2.1742600372750056,36067.5908203125,40756.0,41878.0,100.0,3175432192.0,91215613952.0,360438,e91a616e88278e062635d12bd204679caf879dafc107cb066cf4155c3efe2486
|
| 5 |
+
moe,sdnq,04_underwater_manta,4204,158.60668343293946,2.1726942936019102,36067.27880859375,40756.0,41878.0,100.0,3106979840.0,90012438528.0,359678,add7b1153f552850a2b095afa51998c4e3af4719f3d5c64425105f1af8c7458e
|
| 6 |
+
moe,sdnq,05_factory_battery_cells,4205,158.87487718299963,2.176368180589036,36067.6962890625,40756.0,41878.0,100.0,3307651072.0,87309594624.0,138406,7b6909a2aa49397f574338560cce4a3e47bb1930e55304f897523fa436fe7333
|
benchmark/base/sdnq/summary.json
ADDED
|
@@ -0,0 +1,370 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"cold_generation_seconds": 159.77318301890045,
|
| 3 |
+
"environment": {
|
| 4 |
+
"cuda": "12.8",
|
| 5 |
+
"gpu": "NVIDIA B200",
|
| 6 |
+
"lingbot_moe_expert_backend": "sglang_triton",
|
| 7 |
+
"python": "3.12.3 (main, Aug 14 2025, 17:47:21) [GCC 13.3.0]",
|
| 8 |
+
"torch": "2.8.0+cu128"
|
| 9 |
+
},
|
| 10 |
+
"generation_mean_seconds": 158.85054309077094,
|
| 11 |
+
"hot_generation_mean_seconds": 158.61988310873858,
|
| 12 |
+
"lingbot_video_code_revision": "a2bb04b78edd848500dc27a26e035a95442ae186",
|
| 13 |
+
"load": {
|
| 14 |
+
"nvidia_smi_after": {
|
| 15 |
+
"memory.total": "183359",
|
| 16 |
+
"memory.used": "27698",
|
| 17 |
+
"name": "NVIDIA B200",
|
| 18 |
+
"power.draw": "250.43",
|
| 19 |
+
"temperature.gpu": "32",
|
| 20 |
+
"timestamp": "2026/07/10 08:40:56.535",
|
| 21 |
+
"utilization.gpu": "17"
|
| 22 |
+
},
|
| 23 |
+
"peak_disk_used_bytes": 188465856512.0,
|
| 24 |
+
"peak_gpu_memory_used_mb": 27188.0,
|
| 25 |
+
"peak_gpu_power_watts": 250.52,
|
| 26 |
+
"peak_gpu_utilization_percent": 19.0,
|
| 27 |
+
"peak_process_rss_bytes": 10137681920.0,
|
| 28 |
+
"peak_system_used_bytes": 90347196416.0,
|
| 29 |
+
"sample_count": 131,
|
| 30 |
+
"sample_interval_seconds": 0.25,
|
| 31 |
+
"seconds": 43.81348343298305,
|
| 32 |
+
"torch_allocated_after_mb": 27063.79736328125,
|
| 33 |
+
"torch_peak_allocated_mb": 27063.79736328125,
|
| 34 |
+
"torch_peak_reserved_mb": 27072.0,
|
| 35 |
+
"torch_reserved_after_mb": 27072.0
|
| 36 |
+
},
|
| 37 |
+
"model": "releases/WaveCut-LingBot-Video-MoE-30B-A3B-SDNQ-uint4-static",
|
| 38 |
+
"model_key": "moe",
|
| 39 |
+
"peak_gpu_memory_used_mb": 42504.0,
|
| 40 |
+
"peak_process_rss_bytes": 3307651072.0,
|
| 41 |
+
"peak_system_used_bytes": 138935304192.0,
|
| 42 |
+
"peak_torch_allocated_mb": 36067.6962890625,
|
| 43 |
+
"rows": [
|
| 44 |
+
{
|
| 45 |
+
"batch_cfg": false,
|
| 46 |
+
"cold_generation": true,
|
| 47 |
+
"ffprobe": {
|
| 48 |
+
"programs": [],
|
| 49 |
+
"streams": [
|
| 50 |
+
{
|
| 51 |
+
"codec_name": "h264",
|
| 52 |
+
"duration": "3.041667",
|
| 53 |
+
"height": 480,
|
| 54 |
+
"nb_frames": "73",
|
| 55 |
+
"pix_fmt": "yuv420p",
|
| 56 |
+
"r_frame_rate": "24/1",
|
| 57 |
+
"width": 832
|
| 58 |
+
}
|
| 59 |
+
]
|
| 60 |
+
},
|
| 61 |
+
"fps": 24,
|
| 62 |
+
"generation_seconds": 159.77318301890045,
|
| 63 |
+
"guidance_scale": 3.0,
|
| 64 |
+
"height": 480,
|
| 65 |
+
"model_key": "moe",
|
| 66 |
+
"negative_prompt": "{\"universal_negative\": {\"visual_quality\": [\"low quality\", \"worst quality\", \"blurry\", \"pixelated\", \"jpeg artifacts\", \"low resolution\", \"unstable color\", \"color flicker\", \"underexposed\", \"overexposed\", \"invisible subject\", \"subject hidden in darkness\"], \"artistic_style\": [\"painting\", \"illustration\", \"drawing\", \"cartoon\", \"3d render\", \"cgi\", \"sketch\", \"digital art\"], \"composition_and_content\": [\"text\", \"watermark\", \"signature\", \"logo\", \"subtitles\", \"pillarboxed\", \"side bars\", \"portrait image in landscape frame\"], \"temporal_and_motion_stability\": [\"flickering\", \"jittery\", \"motion blur\", \"temporal inconsistency\", \"warping\", \"morphing\", \"incoherent motion\", \"unnatural movement\", \"static object with sudden jump\", \"frame-to-frame inconsistency\"], \"material_and_structure\": [\"plastic-like glass\", \"unrealistic texture\", \"deformed bottle\", \"liquid freezing improperly\", \"distorted reflections\"]}}",
|
| 67 |
+
"null_cond_clone_zero": false,
|
| 68 |
+
"num_frames": 73,
|
| 69 |
+
"num_inference_steps": 40,
|
| 70 |
+
"nvidia_smi_after": {
|
| 71 |
+
"memory.total": "183359",
|
| 72 |
+
"memory.used": "41878",
|
| 73 |
+
"name": "NVIDIA B200",
|
| 74 |
+
"power.draw": "259.76",
|
| 75 |
+
"temperature.gpu": "42",
|
| 76 |
+
"timestamp": "2026/07/10 08:43:39.302",
|
| 77 |
+
"utilization.gpu": "0"
|
| 78 |
+
},
|
| 79 |
+
"output_file": "results/benchmark/moe/sdnq/videos/01_arctic_research_drone.mp4",
|
| 80 |
+
"output_sha256": "d040042bdb79aacdcafe80a2c4bd621271e3056823a84621c25ee8fd4fff594b",
|
| 81 |
+
"output_size_bytes": 339957,
|
| 82 |
+
"peak_disk_used_bytes": 188465917952.0,
|
| 83 |
+
"peak_gpu_memory_used_mb": 41878.0,
|
| 84 |
+
"peak_gpu_power_watts": 981.74,
|
| 85 |
+
"peak_gpu_utilization_percent": 100.0,
|
| 86 |
+
"peak_process_rss_bytes": 2920222720.0,
|
| 87 |
+
"peak_system_used_bytes": 138935304192.0,
|
| 88 |
+
"prompt": "{\"comprehensive_description\":{\"scene_content_description\":\"A cinematic dusk shot of an Arctic research outpost on a wind-polished ice shelf under green aurora ribbons. A compact orange rover with a roof-mounted lidar mast drives slowly between blue instrument crates and slender weather antennas. Fine snow dust trails behind the tires, and warm light glows from the station windows against the cold cyan landscape. The atmosphere is realistic, high-detail documentary science footage with crisp snow texture, vapor in the air, and physically plausible shadows.\",\"camera_movement_description\":\"The camera begins as a low, smooth drone tracking shot behind the rover, then arcs gently left to reveal the glowing station and aurora-filled sky while keeping the rover dominant in frame.\"},\"camera_info\":{\"color\":\"Cold cyan with warm orange accents\",\"frame_size\":\"Wide\",\"shot_type_angle\":\"Low angle tracking shot\",\"lens_size\":\"Wide lens\",\"composition\":\"Rover in lower center, research station in the background, aurora above\",\"lighting\":\"Soft twilight with window glow\",\"lighting_type\":\"Natural dusk and practical interior lights\"},\"world_knowledge\":[],\"prominent_elements\":[{\"name\":\"orange autonomous research rover\",\"description\":\"A rugged six-wheeled scientific rover with orange body panels, black tires, a lidar mast, and small blinking status lights.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.7s]\",\"action\":\"drives forward slowly across packed snow, tires compressing the surface\"},{\"timestamp\":\"[0.7s - 1.5s]\",\"action\":\"turns slightly left as the lidar mast rotates\"},{\"timestamp\":\"[1.5s - 2.0s]\",\"action\":\"continues toward the lit station while fine snow trails behind\"}],\"location\":\"lower center of the frame\",\"relative_size\":\"dominant\",\"shape_and_color\":\"low rectangular orange body with black wheels and grey sensor mast\",\"texture\":\"matte painted metal, rubber tires, frost on edges\",\"appearance_details\":\"roof lidar, small antennas, narrow headlights, compact cargo rack\",\"relationship\":\"the main moving subject, traveling toward the research outpost\",\"orientation\":\"moving away from the camera and slightly left\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"Arctic research station\",\"description\":\"A modular polar station made of connected white containers with warm yellow light in the windows.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"remains stationary as wind-blown snow passes in front of it\"}],\"location\":\"middle background\",\"relative_size\":\"large\",\"shape_and_color\":\"rectangular white modules with dark seams and glowing windows\",\"texture\":\"frosted metal panels and glass\",\"appearance_details\":\"small stairs, railings, antennas, blue equipment crates nearby\",\"relationship\":\"destination of the rover\",\"orientation\":\"angled three-quarter view\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"aurora ribbons\",\"description\":\"Green aurora curtains stretching across the darkening polar sky.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"shimmer slowly and drift laterally across the sky\"}],\"location\":\"upper half of the frame\",\"relative_size\":\"large\",\"shape_and_color\":\"soft green luminous ribbons\",\"texture\":\"translucent atmospheric glow\",\"appearance_details\":\"layered folds and faint stars behind\",\"relationship\":\"dominates the sky above the station\",\"orientation\":\"horizontal sweeping arcs\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":true,\"number_of_objects\":\"many luminous bands\"}]}",
|
| 89 |
+
"prompt_file": "01_arctic_research_drone.json",
|
| 90 |
+
"prompt_id": "01_arctic_research_drone",
|
| 91 |
+
"raw_frames_file": "results/benchmark/moe/sdnq/raw_frames/01_arctic_research_drone.npy",
|
| 92 |
+
"raw_frames_sha256": "d8c8386e9e015500d0766968a355de2d172b50c2308b2438978d9dc270f815fe",
|
| 93 |
+
"raw_frames_size_bytes": 87459968,
|
| 94 |
+
"sample_count": 553,
|
| 95 |
+
"sample_interval_seconds": 0.25,
|
| 96 |
+
"seconds_per_frame": 2.188673739984938,
|
| 97 |
+
"seed": 4201,
|
| 98 |
+
"shift": 3.0,
|
| 99 |
+
"torch_allocated_after_mb": 27094.3193359375,
|
| 100 |
+
"torch_peak_allocated_mb": 36067.29833984375,
|
| 101 |
+
"torch_peak_reserved_mb": 40756.0,
|
| 102 |
+
"torch_reserved_after_mb": 40756.0,
|
| 103 |
+
"variant": "sdnq",
|
| 104 |
+
"width": 832
|
| 105 |
+
},
|
| 106 |
+
{
|
| 107 |
+
"batch_cfg": false,
|
| 108 |
+
"cold_generation": false,
|
| 109 |
+
"ffprobe": {
|
| 110 |
+
"programs": [],
|
| 111 |
+
"streams": [
|
| 112 |
+
{
|
| 113 |
+
"codec_name": "h264",
|
| 114 |
+
"duration": "3.041667",
|
| 115 |
+
"height": 480,
|
| 116 |
+
"nb_frames": "73",
|
| 117 |
+
"pix_fmt": "yuv420p",
|
| 118 |
+
"r_frame_rate": "24/1",
|
| 119 |
+
"width": 832
|
| 120 |
+
}
|
| 121 |
+
]
|
| 122 |
+
},
|
| 123 |
+
"fps": 24,
|
| 124 |
+
"generation_seconds": 158.2769890979398,
|
| 125 |
+
"guidance_scale": 3.0,
|
| 126 |
+
"height": 480,
|
| 127 |
+
"model_key": "moe",
|
| 128 |
+
"negative_prompt": "{\"universal_negative\": {\"visual_quality\": [\"low quality\", \"worst quality\", \"blurry\", \"pixelated\", \"jpeg artifacts\", \"low resolution\", \"unstable color\", \"color flicker\", \"underexposed\", \"overexposed\", \"invisible subject\", \"subject hidden in darkness\"], \"artistic_style\": [\"painting\", \"illustration\", \"drawing\", \"cartoon\", \"3d render\", \"cgi\", \"sketch\", \"digital art\"], \"composition_and_content\": [\"text\", \"watermark\", \"signature\", \"logo\", \"subtitles\", \"pillarboxed\", \"side bars\", \"portrait image in landscape frame\"], \"temporal_and_motion_stability\": [\"flickering\", \"jittery\", \"motion blur\", \"temporal inconsistency\", \"warping\", \"morphing\", \"incoherent motion\", \"unnatural movement\", \"static object with sudden jump\", \"frame-to-frame inconsistency\"], \"material_and_structure\": [\"plastic-like glass\", \"unrealistic texture\", \"deformed bottle\", \"liquid freezing improperly\", \"distorted reflections\"]}}",
|
| 129 |
+
"null_cond_clone_zero": false,
|
| 130 |
+
"num_frames": 73,
|
| 131 |
+
"num_inference_steps": 40,
|
| 132 |
+
"nvidia_smi_after": {
|
| 133 |
+
"memory.total": "183359",
|
| 134 |
+
"memory.used": "41878",
|
| 135 |
+
"name": "NVIDIA B200",
|
| 136 |
+
"power.draw": "260.53",
|
| 137 |
+
"temperature.gpu": "43",
|
| 138 |
+
"timestamp": "2026/07/10 08:46:20.584",
|
| 139 |
+
"utilization.gpu": "0"
|
| 140 |
+
},
|
| 141 |
+
"output_file": "results/benchmark/moe/sdnq/videos/02_macro_pastry_glaze.mp4",
|
| 142 |
+
"output_sha256": "bcb2caabb9e29427019ef3dcc7790a89910c1b570e65f85bac3f8a95dc1912db",
|
| 143 |
+
"output_size_bytes": 367773,
|
| 144 |
+
"peak_disk_used_bytes": 188554317824.0,
|
| 145 |
+
"peak_gpu_memory_used_mb": 42504.0,
|
| 146 |
+
"peak_gpu_power_watts": 976.17,
|
| 147 |
+
"peak_gpu_utilization_percent": 100.0,
|
| 148 |
+
"peak_process_rss_bytes": 3105587200.0,
|
| 149 |
+
"peak_system_used_bytes": 90820800512.0,
|
| 150 |
+
"prompt": "{\"comprehensive_description\":{\"scene_content_description\":\"An extreme macro food cinematography shot in a professional pastry kitchen. A glossy dark chocolate dome dessert sits on a rotating metal turntable. A thin stream of amber caramel glaze pours from above, flowing over the curved surface in slow ribbons and revealing tiny reflections from softbox lights. Crushed pistachio dust and edible gold flakes sit around the base on a matte black plate. The look is photorealistic, tactile, high-speed macro video with shallow depth of field and no text.\",\"camera_movement_description\":\"The camera holds a close macro angle with a subtle push-in while the dessert rotates clockwise on the turntable.\"},\"camera_info\":{\"color\":\"Deep brown, amber, green pistachio, gold highlights\",\"frame_size\":\"Extreme close up\",\"shot_type_angle\":\"Slight high angle\",\"lens_size\":\"Macro lens\",\"composition\":\"Dessert centered with glaze entering from top\",\"lighting\":\"Large softbox reflections\",\"lighting_type\":\"Controlled studio kitchen lighting\"},\"world_knowledge\":[],\"prominent_elements\":[{\"name\":\"chocolate dome dessert\",\"description\":\"A small mirror-glazed chocolate dome on a black plate, surrounded by pistachio dust and gold flakes.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"rotates slowly clockwise on a metal turntable\"}],\"location\":\"center of the frame\",\"relative_size\":\"dominant\",\"shape_and_color\":\"hemispherical dark chocolate dome on round black plate\",\"texture\":\"mirror-gloss surface with smooth curved reflections\",\"appearance_details\":\"tiny gold flakes, green pistachio crumbs, clean pastry plating\",\"relationship\":\"receives the caramel glaze\",\"orientation\":\"upright on the turntable\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"caramel glaze stream\",\"description\":\"A viscous amber caramel stream pouring from above and spreading over the dessert.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.8s]\",\"action\":\"falls as a narrow glossy stream onto the crown of the dome\"},{\"timestamp\":\"[0.8s - 1.6s]\",\"action\":\"forms slow ribbons that slide down the curved chocolate surface\"},{\"timestamp\":\"[1.6s - 2.0s]\",\"action\":\"collects at the base in a thin shiny ring\"}],\"location\":\"entering from top center and flowing over the dessert\",\"relative_size\":\"medium\",\"shape_and_color\":\"thin amber liquid ribbons\",\"texture\":\"viscous, glossy, translucent caramel\",\"appearance_details\":\"bright highlights and tiny bubbles\",\"relationship\":\"coats the chocolate dome\",\"orientation\":\"vertical stream becoming downward trails\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"}]}",
|
| 151 |
+
"prompt_file": "02_macro_pastry_glaze.json",
|
| 152 |
+
"prompt_id": "02_macro_pastry_glaze",
|
| 153 |
+
"raw_frames_file": "results/benchmark/moe/sdnq/raw_frames/02_macro_pastry_glaze.npy",
|
| 154 |
+
"raw_frames_sha256": "0c455e53a36b1ca3a8e933229529ff2541faf889e504d4527ad9795e8083ae6e",
|
| 155 |
+
"raw_frames_size_bytes": 87459968,
|
| 156 |
+
"sample_count": 538,
|
| 157 |
+
"sample_interval_seconds": 0.25,
|
| 158 |
+
"seconds_per_frame": 2.1681779328484905,
|
| 159 |
+
"seed": 4202,
|
| 160 |
+
"shift": 3.0,
|
| 161 |
+
"torch_allocated_after_mb": 27094.3193359375,
|
| 162 |
+
"torch_peak_allocated_mb": 36065.68798828125,
|
| 163 |
+
"torch_peak_reserved_mb": 40756.0,
|
| 164 |
+
"torch_reserved_after_mb": 40756.0,
|
| 165 |
+
"variant": "sdnq",
|
| 166 |
+
"width": 832
|
| 167 |
+
},
|
| 168 |
+
{
|
| 169 |
+
"batch_cfg": false,
|
| 170 |
+
"cold_generation": false,
|
| 171 |
+
"ffprobe": {
|
| 172 |
+
"programs": [],
|
| 173 |
+
"streams": [
|
| 174 |
+
{
|
| 175 |
+
"codec_name": "h264",
|
| 176 |
+
"duration": "3.041667",
|
| 177 |
+
"height": 480,
|
| 178 |
+
"nb_frames": "73",
|
| 179 |
+
"pix_fmt": "yuv420p",
|
| 180 |
+
"r_frame_rate": "24/1",
|
| 181 |
+
"width": 832
|
| 182 |
+
}
|
| 183 |
+
]
|
| 184 |
+
},
|
| 185 |
+
"fps": 24,
|
| 186 |
+
"generation_seconds": 158.72098272107542,
|
| 187 |
+
"guidance_scale": 3.0,
|
| 188 |
+
"height": 480,
|
| 189 |
+
"model_key": "moe",
|
| 190 |
+
"negative_prompt": "{\"universal_negative\": {\"visual_quality\": [\"low quality\", \"worst quality\", \"blurry\", \"pixelated\", \"jpeg artifacts\", \"low resolution\", \"unstable color\", \"color flicker\", \"underexposed\", \"overexposed\", \"invisible subject\", \"subject hidden in darkness\"], \"artistic_style\": [\"painting\", \"illustration\", \"drawing\", \"cartoon\", \"3d render\", \"cgi\", \"sketch\", \"digital art\"], \"composition_and_content\": [\"text\", \"watermark\", \"signature\", \"logo\", \"subtitles\", \"pillarboxed\", \"side bars\", \"portrait image in landscape frame\"], \"temporal_and_motion_stability\": [\"flickering\", \"jittery\", \"motion blur\", \"temporal inconsistency\", \"warping\", \"morphing\", \"incoherent motion\", \"unnatural movement\", \"static object with sudden jump\", \"frame-to-frame inconsistency\"], \"material_and_structure\": [\"plastic-like glass\", \"unrealistic texture\", \"deformed bottle\", \"liquid freezing improperly\", \"distorted reflections\"]}}",
|
| 191 |
+
"null_cond_clone_zero": false,
|
| 192 |
+
"num_frames": 73,
|
| 193 |
+
"num_inference_steps": 40,
|
| 194 |
+
"nvidia_smi_after": {
|
| 195 |
+
"memory.total": "183359",
|
| 196 |
+
"memory.used": "41878",
|
| 197 |
+
"name": "NVIDIA B200",
|
| 198 |
+
"power.draw": "260.26",
|
| 199 |
+
"temperature.gpu": "43",
|
| 200 |
+
"timestamp": "2026/07/10 08:49:02.315",
|
| 201 |
+
"utilization.gpu": "0"
|
| 202 |
+
},
|
| 203 |
+
"output_file": "results/benchmark/moe/sdnq/videos/03_rain_market_lanterns.mp4",
|
| 204 |
+
"output_sha256": "e91a616e88278e062635d12bd204679caf879dafc107cb066cf4155c3efe2486",
|
| 205 |
+
"output_size_bytes": 360438,
|
| 206 |
+
"peak_disk_used_bytes": 183805046784.0,
|
| 207 |
+
"peak_gpu_memory_used_mb": 41878.0,
|
| 208 |
+
"peak_gpu_power_watts": 975.4,
|
| 209 |
+
"peak_gpu_utilization_percent": 100.0,
|
| 210 |
+
"peak_process_rss_bytes": 3175432192.0,
|
| 211 |
+
"peak_system_used_bytes": 91215613952.0,
|
| 212 |
+
"prompt": "{\"comprehensive_description\":{\"scene_content_description\":\"A lively rainy night market in a narrow old-city street, filmed like naturalistic cinema. Red paper lanterns hang from wooden stalls, steam rises from bowls of soup, and wet cobblestones mirror orange and teal reflections. A vendor in a dark apron slides a steaming bowl across a counter to a traveler in a blue raincoat. Umbrellas pass in the foreground, partially occluding the scene for a moment. The mood is warm, humid, detailed, and grounded, with no readable signs or subtitles.\",\"camera_movement_description\":\"The camera performs a slow lateral dolly from left to right at counter height, with foreground umbrellas briefly crossing the lens.\"},\"camera_info\":{\"color\":\"Warm lantern orange, teal rain reflections, red accents\",\"frame_size\":\"Medium wide\",\"shot_type_angle\":\"Eye level\",\"lens_size\":\"Medium lens\",\"composition\":\"Vendor stall centered, traveler on right, passing umbrellas in foreground\",\"lighting\":\"Lantern light mixed with rain reflections\",\"lighting_type\":\"Night practical lighting\"},\"world_knowledge\":[],\"prominent_elements\":[{\"name\":\"food vendor\",\"description\":\"A middle-aged street vendor wearing a dark apron and working behind a steaming wooden food stall.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.8s]\",\"action\":\"lifts a white ceramic bowl from a steaming pot\"},{\"timestamp\":\"[0.8s - 1.6s]\",\"action\":\"slides the bowl carefully across the counter\"},{\"timestamp\":\"[1.6s - 2.0s]\",\"action\":\"wipes condensation from the counter with a cloth\"}],\"location\":\"center-left behind the stall counter\",\"relative_size\":\"large\",\"shape_and_color\":\"human figure in dark apron under warm lantern light\",\"texture\":\"wet fabric, skin highlights, wood grain around the stall\",\"appearance_details\":\"rolled sleeves, focused expression, steam around hands\",\"relationship\":\"serves food to the traveler\",\"orientation\":\"facing right toward the customer\",\"pose\":\"leaning slightly forward over the counter\",\"expression\":\"focused and calm\",\"clothing\":\"dark apron over a muted shirt\",\"gender\":\"\",\"skin_tone_and_texture\":\"natural skin with rain-lit highlights\"},{\"name\":\"traveler in blue raincoat\",\"description\":\"A traveler wearing a blue hooded raincoat, waiting at the stall with wet sleeves and a folded umbrella.\",\"actions\":[{\"timestamp\":\"[0.0s - 1.2s]\",\"action\":\"waits with hands near the counter, watching the bowl\"},{\"timestamp\":\"[1.2s - 2.0s]\",\"action\":\"reaches forward to receive the steaming bowl\"}],\"location\":\"right side of the frame\",\"relative_size\":\"medium\",\"shape_and_color\":\"blue hooded raincoat silhouette\",\"texture\":\"water beads on waterproof fabric\",\"appearance_details\":\"hood up, folded umbrella tucked under one arm\",\"relationship\":\"customer receiving food from the vendor\",\"orientation\":\"facing left toward the stall\",\"pose\":\"standing close to the counter\",\"expression\":\"expectant and tired\",\"clothing\":\"blue raincoat\",\"gender\":\"\",\"skin_tone_and_texture\":\"\"},{\"name\":\"rain and lantern reflections\",\"description\":\"Falling rain, steam, and glowing reflections on wet cobblestones.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"rain falls continuously while reflections ripple on the street\"}],\"location\":\"foreground and background throughout the frame\",\"relative_size\":\"large\",\"shape_and_color\":\"thin rain streaks, orange and teal reflected pools\",\"texture\":\"wet stone, vapor, glossy water\",\"appearance_details\":\"soft steam plumes, umbrellas passing close to lens\",\"relationship\":\"sets the atmosphere around the market\",\"orientation\":\"vertical rainfall and horizontal street reflections\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":true,\"number_of_objects\":\"many raindrops and reflections\"}]}",
|
| 213 |
+
"prompt_file": "03_rain_market_lanterns.json",
|
| 214 |
+
"prompt_id": "03_rain_market_lanterns",
|
| 215 |
+
"raw_frames_file": "results/benchmark/moe/sdnq/raw_frames/03_rain_market_lanterns.npy",
|
| 216 |
+
"raw_frames_sha256": "535b3669bd1fc07990872df951b2544288ac4e4554000d9c4fb705a24dd33ed3",
|
| 217 |
+
"raw_frames_size_bytes": 87459968,
|
| 218 |
+
"sample_count": 549,
|
| 219 |
+
"sample_interval_seconds": 0.25,
|
| 220 |
+
"seconds_per_frame": 2.1742600372750056,
|
| 221 |
+
"seed": 4203,
|
| 222 |
+
"shift": 3.0,
|
| 223 |
+
"torch_allocated_after_mb": 27094.3193359375,
|
| 224 |
+
"torch_peak_allocated_mb": 36067.5908203125,
|
| 225 |
+
"torch_peak_reserved_mb": 40756.0,
|
| 226 |
+
"torch_reserved_after_mb": 40756.0,
|
| 227 |
+
"variant": "sdnq",
|
| 228 |
+
"width": 832
|
| 229 |
+
},
|
| 230 |
+
{
|
| 231 |
+
"batch_cfg": false,
|
| 232 |
+
"cold_generation": false,
|
| 233 |
+
"ffprobe": {
|
| 234 |
+
"programs": [],
|
| 235 |
+
"streams": [
|
| 236 |
+
{
|
| 237 |
+
"codec_name": "h264",
|
| 238 |
+
"duration": "3.041667",
|
| 239 |
+
"height": 480,
|
| 240 |
+
"nb_frames": "73",
|
| 241 |
+
"pix_fmt": "yuv420p",
|
| 242 |
+
"r_frame_rate": "24/1",
|
| 243 |
+
"width": 832
|
| 244 |
+
}
|
| 245 |
+
]
|
| 246 |
+
},
|
| 247 |
+
"fps": 24,
|
| 248 |
+
"generation_seconds": 158.60668343293946,
|
| 249 |
+
"guidance_scale": 3.0,
|
| 250 |
+
"height": 480,
|
| 251 |
+
"model_key": "moe",
|
| 252 |
+
"negative_prompt": "{\"universal_negative\": {\"visual_quality\": [\"low quality\", \"worst quality\", \"blurry\", \"pixelated\", \"jpeg artifacts\", \"low resolution\", \"unstable color\", \"color flicker\", \"underexposed\", \"overexposed\", \"invisible subject\", \"subject hidden in darkness\"], \"artistic_style\": [\"painting\", \"illustration\", \"drawing\", \"cartoon\", \"3d render\", \"cgi\", \"sketch\", \"digital art\"], \"composition_and_content\": [\"text\", \"watermark\", \"signature\", \"logo\", \"subtitles\", \"pillarboxed\", \"side bars\", \"portrait image in landscape frame\"], \"temporal_and_motion_stability\": [\"flickering\", \"jittery\", \"motion blur\", \"temporal inconsistency\", \"warping\", \"morphing\", \"incoherent motion\", \"unnatural movement\", \"static object with sudden jump\", \"frame-to-frame inconsistency\"], \"material_and_structure\": [\"plastic-like glass\", \"unrealistic texture\", \"deformed bottle\", \"liquid freezing improperly\", \"distorted reflections\"]}}",
|
| 253 |
+
"null_cond_clone_zero": false,
|
| 254 |
+
"num_frames": 73,
|
| 255 |
+
"num_inference_steps": 40,
|
| 256 |
+
"nvidia_smi_after": {
|
| 257 |
+
"memory.total": "183359",
|
| 258 |
+
"memory.used": "41878",
|
| 259 |
+
"name": "NVIDIA B200",
|
| 260 |
+
"power.draw": "260.36",
|
| 261 |
+
"temperature.gpu": "43",
|
| 262 |
+
"timestamp": "2026/07/10 08:51:43.973",
|
| 263 |
+
"utilization.gpu": "0"
|
| 264 |
+
},
|
| 265 |
+
"output_file": "results/benchmark/moe/sdnq/videos/04_underwater_manta.mp4",
|
| 266 |
+
"output_sha256": "add7b1153f552850a2b095afa51998c4e3af4719f3d5c64425105f1af8c7458e",
|
| 267 |
+
"output_size_bytes": 359678,
|
| 268 |
+
"peak_disk_used_bytes": 183893323776.0,
|
| 269 |
+
"peak_gpu_memory_used_mb": 41878.0,
|
| 270 |
+
"peak_gpu_power_watts": 981.58,
|
| 271 |
+
"peak_gpu_utilization_percent": 100.0,
|
| 272 |
+
"peak_process_rss_bytes": 3106979840.0,
|
| 273 |
+
"peak_system_used_bytes": 90012438528.0,
|
| 274 |
+
"prompt": "{\"comprehensive_description\":{\"scene_content_description\":\"A tranquil underwater wildlife scene in clear tropical water. A large manta ray glides above a colorful coral garden while a scuba diver hovers several meters behind, holding a small camera rig. Sunbeams ripple through the surface, schools of tiny silver fish split and rejoin around the manta, and suspended particles drift slowly. The image should feel like realistic underwater documentary footage with natural motion, soft blue-green color, and no fantasy elements.\",\"camera_movement_description\":\"The camera tracks smoothly beside the manta ray from a slightly lower angle, drifting forward with the current.\"},\"camera_info\":{\"color\":\"Blue-green water with coral reds and yellows\",\"frame_size\":\"Wide\",\"shot_type_angle\":\"Slight low angle\",\"lens_size\":\"Wide underwater lens\",\"composition\":\"Manta ray crossing the upper center, coral below, diver in background\",\"lighting\":\"Dappled sunlight\",\"lighting_type\":\"Natural underwater daylight\"},\"world_knowledge\":[],\"prominent_elements\":[{\"name\":\"manta ray\",\"description\":\"A large manta ray with broad triangular fins, dark top surface, and pale underside.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.8s]\",\"action\":\"glides from left to right with slow wing-like fin movement\"},{\"timestamp\":\"[0.8s - 1.5s]\",\"action\":\"tilts slightly upward as sunbeams cross its back\"},{\"timestamp\":\"[1.5s - 2.0s]\",\"action\":\"continues forward while small fish scatter around it\"}],\"location\":\"upper center moving toward right\",\"relative_size\":\"dominant\",\"shape_and_color\":\"wide diamond-like silhouette, dark grey top, pale underside\",\"texture\":\"smooth skin with subtle mottling\",\"appearance_details\":\"cephalic fins near mouth, long tail trailing behind\",\"relationship\":\"main wildlife subject of the video\",\"orientation\":\"moving left to right and slightly upward\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"scuba diver\",\"description\":\"A diver in black wetsuit and fins, carrying a small underwater camera rig.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"hovers calmly in the background, exhaling small bubble streams\"}],\"location\":\"mid-background behind the manta ray\",\"relative_size\":\"small\",\"shape_and_color\":\"human figure in black wetsuit with silver air tank\",\"texture\":\"neoprene suit, metal tank, glass mask\",\"appearance_details\":\"fins, mask, regulator, compact camera lights\",\"relationship\":\"observes the manta ray without touching it\",\"orientation\":\"facing the manta ray\",\"pose\":\"horizontal hover\",\"expression\":\"\",\"clothing\":\"black scuba gear\",\"gender\":\"\",\"skin_tone_and_texture\":\"\"},{\"name\":\"coral reef and fish\",\"description\":\"A colorful coral garden with small silver fish schooling above it.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"fish swirl and split around the manta while coral remains fixed\"}],\"location\":\"lower half of the frame\",\"relative_size\":\"large\",\"shape_and_color\":\"branching corals in red, yellow, and muted purple; silver fish clusters\",\"texture\":\"rough coral, shimmering fish scales\",\"appearance_details\":\"sand patches, sea fans, drifting particles\",\"relationship\":\"environment below the manta ray\",\"orientation\":\"reef spreads horizontally across the bottom\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":true,\"number_of_objects\":\"many fish and coral structures\"}]}",
|
| 275 |
+
"prompt_file": "04_underwater_manta.json",
|
| 276 |
+
"prompt_id": "04_underwater_manta",
|
| 277 |
+
"raw_frames_file": "results/benchmark/moe/sdnq/raw_frames/04_underwater_manta.npy",
|
| 278 |
+
"raw_frames_sha256": "7a51833e767e2dffec08ef8a47ab57a5541ad34cdedc30553362369cd287f2c9",
|
| 279 |
+
"raw_frames_size_bytes": 87459968,
|
| 280 |
+
"sample_count": 555,
|
| 281 |
+
"sample_interval_seconds": 0.25,
|
| 282 |
+
"seconds_per_frame": 2.1726942936019102,
|
| 283 |
+
"seed": 4204,
|
| 284 |
+
"shift": 3.0,
|
| 285 |
+
"torch_allocated_after_mb": 27094.3193359375,
|
| 286 |
+
"torch_peak_allocated_mb": 36067.27880859375,
|
| 287 |
+
"torch_peak_reserved_mb": 40756.0,
|
| 288 |
+
"torch_reserved_after_mb": 40756.0,
|
| 289 |
+
"variant": "sdnq",
|
| 290 |
+
"width": 832
|
| 291 |
+
},
|
| 292 |
+
{
|
| 293 |
+
"batch_cfg": false,
|
| 294 |
+
"cold_generation": false,
|
| 295 |
+
"ffprobe": {
|
| 296 |
+
"programs": [],
|
| 297 |
+
"streams": [
|
| 298 |
+
{
|
| 299 |
+
"codec_name": "h264",
|
| 300 |
+
"duration": "3.041667",
|
| 301 |
+
"height": 480,
|
| 302 |
+
"nb_frames": "73",
|
| 303 |
+
"pix_fmt": "yuv420p",
|
| 304 |
+
"r_frame_rate": "24/1",
|
| 305 |
+
"width": 832
|
| 306 |
+
}
|
| 307 |
+
]
|
| 308 |
+
},
|
| 309 |
+
"fps": 24,
|
| 310 |
+
"generation_seconds": 158.87487718299963,
|
| 311 |
+
"guidance_scale": 3.0,
|
| 312 |
+
"height": 480,
|
| 313 |
+
"model_key": "moe",
|
| 314 |
+
"negative_prompt": "{\"universal_negative\": {\"visual_quality\": [\"low quality\", \"worst quality\", \"blurry\", \"pixelated\", \"jpeg artifacts\", \"low resolution\", \"unstable color\", \"color flicker\", \"underexposed\", \"overexposed\", \"invisible subject\", \"subject hidden in darkness\"], \"artistic_style\": [\"painting\", \"illustration\", \"drawing\", \"cartoon\", \"3d render\", \"cgi\", \"sketch\", \"digital art\"], \"composition_and_content\": [\"text\", \"watermark\", \"signature\", \"logo\", \"subtitles\", \"pillarboxed\", \"side bars\", \"portrait image in landscape frame\"], \"temporal_and_motion_stability\": [\"flickering\", \"jittery\", \"motion blur\", \"temporal inconsistency\", \"warping\", \"morphing\", \"incoherent motion\", \"unnatural movement\", \"static object with sudden jump\", \"frame-to-frame inconsistency\"], \"material_and_structure\": [\"plastic-like glass\", \"unrealistic texture\", \"deformed bottle\", \"liquid freezing improperly\", \"distorted reflections\"]}}",
|
| 315 |
+
"null_cond_clone_zero": false,
|
| 316 |
+
"num_frames": 73,
|
| 317 |
+
"num_inference_steps": 40,
|
| 318 |
+
"nvidia_smi_after": {
|
| 319 |
+
"memory.total": "183359",
|
| 320 |
+
"memory.used": "41878",
|
| 321 |
+
"name": "NVIDIA B200",
|
| 322 |
+
"power.draw": "260.35",
|
| 323 |
+
"temperature.gpu": "43",
|
| 324 |
+
"timestamp": "2026/07/10 08:54:25.766",
|
| 325 |
+
"utilization.gpu": "0"
|
| 326 |
+
},
|
| 327 |
+
"output_file": "results/benchmark/moe/sdnq/videos/05_factory_battery_cells.mp4",
|
| 328 |
+
"output_sha256": "7b6909a2aa49397f574338560cce4a3e47bb1930e55304f897523fa436fe7333",
|
| 329 |
+
"output_size_bytes": 138406,
|
| 330 |
+
"peak_disk_used_bytes": 183981420544.0,
|
| 331 |
+
"peak_gpu_memory_used_mb": 41878.0,
|
| 332 |
+
"peak_gpu_power_watts": 981.13,
|
| 333 |
+
"peak_gpu_utilization_percent": 100.0,
|
| 334 |
+
"peak_process_rss_bytes": 3307651072.0,
|
| 335 |
+
"peak_system_used_bytes": 87309594624.0,
|
| 336 |
+
"prompt": "{\"comprehensive_description\":{\"scene_content_description\":\"A clean high-tech factory line assembling transparent solid-state battery cells. Two white robotic arms move with precise synchronized motion over a brushed steel conveyor. One arm lowers a translucent rectangular cell into a copper test fixture while the other arm scans it with a blue light bar. Tiny reflections slide across glass safety panels, and status lights pulse softly. The scene should look like realistic industrial automation footage, not CGI, with crisp metal surfaces and controlled motion.\",\"camera_movement_description\":\"The camera is locked off in a medium-wide three-quarter view with a subtle mechanical vibration, emphasizing repeatable robotic motion.\"},\"camera_info\":{\"color\":\"White, brushed steel, copper, blue inspection light\",\"frame_size\":\"Medium wide\",\"shot_type_angle\":\"Three-quarter eye-level industrial view\",\"lens_size\":\"Medium lens\",\"composition\":\"Two robotic arms framing the battery cell at center\",\"lighting\":\"Even overhead factory lighting with blue scanner highlight\",\"lighting_type\":\"Artificial industrial lighting\"},\"world_knowledge\":[],\"prominent_elements\":[{\"name\":\"left robotic arm\",\"description\":\"A white six-axis robotic arm with a vacuum gripper holding a transparent rectangular battery cell.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.7s]\",\"action\":\"moves downward from upper left with the transparent cell held steady\"},{\"timestamp\":\"[0.7s - 1.4s]\",\"action\":\"places the cell into a copper test fixture\"},{\"timestamp\":\"[1.4s - 2.0s]\",\"action\":\"releases the cell and retracts slightly upward\"}],\"location\":\"left side moving toward center\",\"relative_size\":\"large\",\"shape_and_color\":\"white articulated segments with black joints\",\"texture\":\"smooth painted metal and rubber vacuum cups\",\"appearance_details\":\"visible cable routing, small green status LED\",\"relationship\":\"positions the battery cell for testing\",\"orientation\":\"angled downward toward the center fixture\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"right robotic scanner arm\",\"description\":\"A second white robotic arm carrying a rectangular blue inspection light bar.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.9s]\",\"action\":\"waits above the copper fixture with blue scanner light dim\"},{\"timestamp\":\"[0.9s - 1.6s]\",\"action\":\"sweeps the blue light bar across the transparent cell\"},{\"timestamp\":\"[1.6s - 2.0s]\",\"action\":\"pauses as the scanner light pulses once\"}],\"location\":\"right side above the conveyor\",\"relative_size\":\"large\",\"shape_and_color\":\"white arm with glowing blue rectangular scanner\",\"texture\":\"smooth metal casing and glass scanner cover\",\"appearance_details\":\"blue light strip, black joints, compact sensor module\",\"relationship\":\"inspects the cell after placement\",\"orientation\":\"angled left toward the battery cell\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"transparent battery cell and copper fixture\",\"description\":\"A clear rectangular solid-state cell seated into a copper test fixture on a steel conveyor.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.7s]\",\"action\":\"moves with the left arm above the fixture\"},{\"timestamp\":\"[0.7s - 1.4s]\",\"action\":\"is lowered precisely into the fixture\"},{\"timestamp\":\"[1.4s - 2.0s]\",\"action\":\"remains stationary while blue inspection light passes over it\"}],\"location\":\"center of the frame on the conveyor\",\"relative_size\":\"medium\",\"shape_and_color\":\"transparent rectangle with faint internal layers, copper fixture\",\"texture\":\"clear glass-like cell, brushed copper, polished steel\",\"appearance_details\":\"thin internal laminate layers, small alignment pins, clean conveyor surface\",\"relationship\":\"object being assembled and inspected\",\"orientation\":\"horizontal in the fixture\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"}]}",
|
| 337 |
+
"prompt_file": "05_factory_battery_cells.json",
|
| 338 |
+
"prompt_id": "05_factory_battery_cells",
|
| 339 |
+
"raw_frames_file": "results/benchmark/moe/sdnq/raw_frames/05_factory_battery_cells.npy",
|
| 340 |
+
"raw_frames_sha256": "2a8b445ef18edfe1789e3d4b779b9a8eb290055605f24b4e7cef890f927a3e5e",
|
| 341 |
+
"raw_frames_size_bytes": 87459968,
|
| 342 |
+
"sample_count": 556,
|
| 343 |
+
"sample_interval_seconds": 0.25,
|
| 344 |
+
"seconds_per_frame": 2.176368180589036,
|
| 345 |
+
"seed": 4205,
|
| 346 |
+
"shift": 3.0,
|
| 347 |
+
"torch_allocated_after_mb": 27094.3193359375,
|
| 348 |
+
"torch_peak_allocated_mb": 36067.6962890625,
|
| 349 |
+
"torch_peak_reserved_mb": 40756.0,
|
| 350 |
+
"torch_reserved_after_mb": 40756.0,
|
| 351 |
+
"variant": "sdnq",
|
| 352 |
+
"width": 832
|
| 353 |
+
}
|
| 354 |
+
],
|
| 355 |
+
"sdnq_revision": "d841c383ff7be38728d4df829e17af4f15d4fd66",
|
| 356 |
+
"settings": {
|
| 357 |
+
"batch_cfg": false,
|
| 358 |
+
"fps": 24,
|
| 359 |
+
"guidance_scale": 3.0,
|
| 360 |
+
"height": 480,
|
| 361 |
+
"null_cond_clone_zero": false,
|
| 362 |
+
"num_frames": 73,
|
| 363 |
+
"num_inference_steps": 40,
|
| 364 |
+
"shift": 3.0,
|
| 365 |
+
"width": 832
|
| 366 |
+
},
|
| 367 |
+
"source_repo_id": "robbyant/lingbot-video-moe-30b-a3b",
|
| 368 |
+
"source_revision": "f2e538f64afe00cc4ae674db2aeb52e2945edfd5",
|
| 369 |
+
"variant": "sdnq"
|
| 370 |
+
}
|
benchmark/coverage/refiner.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
benchmark/coverage/transformer.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
benchmark/environment/environment.json
ADDED
|
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"accelerate": "1.14.0",
|
| 3 |
+
"cuda": "12.8",
|
| 4 |
+
"cuda_available": true,
|
| 5 |
+
"cuda_device": "NVIDIA B200",
|
| 6 |
+
"diffusers": "0.39.0",
|
| 7 |
+
"huggingface_hub": "1.23.0",
|
| 8 |
+
"platform": "Linux-6.8.0-124-generic-x86_64-with-glibc2.39",
|
| 9 |
+
"python": "3.12.3",
|
| 10 |
+
"safetensors": "0.8.0",
|
| 11 |
+
"sdnq_file": "/workspace/lingbot-video-sdnq/sources/sdnq/src/sdnq/__init__.py",
|
| 12 |
+
"torch": "2.8.0+cu128",
|
| 13 |
+
"transformers": "5.8.1"
|
| 14 |
+
}
|
benchmark/environment/nvidia-smi-q.txt
ADDED
|
@@ -0,0 +1,258 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
|
| 2 |
+
==============NVSMI LOG==============
|
| 3 |
+
|
| 4 |
+
Timestamp : Fri Jul 10 07:54:56 2026
|
| 5 |
+
Driver Version : 595.71.05
|
| 6 |
+
CUDA Version : 13.2
|
| 7 |
+
|
| 8 |
+
Attached GPUs : 1
|
| 9 |
+
GPU 00000000:17:00.0
|
| 10 |
+
Product Name : NVIDIA B200
|
| 11 |
+
Product Brand : NVIDIA
|
| 12 |
+
Product Architecture : Blackwell
|
| 13 |
+
Display Mode : Requested functionality has been deprecated
|
| 14 |
+
Display Attached : Yes
|
| 15 |
+
Display Active : Disabled
|
| 16 |
+
Persistence Mode : Enabled
|
| 17 |
+
Addressing Mode : None
|
| 18 |
+
MIG Mode
|
| 19 |
+
Current : Disabled
|
| 20 |
+
Pending : Disabled
|
| 21 |
+
Accounting Mode : Disabled
|
| 22 |
+
Accounting Mode Buffer Size : 4000
|
| 23 |
+
Driver Model
|
| 24 |
+
Current : N/A
|
| 25 |
+
Pending : N/A
|
| 26 |
+
Serial Number : 1650526070674
|
| 27 |
+
GPU UUID : GPU-894b3998-1968-8750-364f-2fd6553c0678
|
| 28 |
+
GPU PDI : 0x584812bd84a56e07
|
| 29 |
+
Minor Number : 0
|
| 30 |
+
VBIOS Version : 97.00.E4.00.11
|
| 31 |
+
MultiGPU Board : No
|
| 32 |
+
Board ID : 0x1700
|
| 33 |
+
Board Part Number : 692-2G525-0220-500
|
| 34 |
+
GPU Part Number : 2901-886-A1
|
| 35 |
+
FRU Part Number : N/A
|
| 36 |
+
Platform Info
|
| 37 |
+
Chassis Serial Number :
|
| 38 |
+
Slot Number : N/A
|
| 39 |
+
Tray Index : N/A
|
| 40 |
+
Host ID : 1
|
| 41 |
+
Peer Type : Switch Connected
|
| 42 |
+
Module Id : 6
|
| 43 |
+
GPU Fabric GUID : 0x78375a0003e23aac
|
| 44 |
+
Inforom Version
|
| 45 |
+
Image Version : G525.0220.00.03
|
| 46 |
+
OEM Object : 2.1
|
| 47 |
+
ECC Object : 7.16
|
| 48 |
+
Power Management Object : N/A
|
| 49 |
+
Inforom BBX Object Flush
|
| 50 |
+
Latest Timestamp : 2026/07/09 21:45:51.022
|
| 51 |
+
Latest Duration : 59141 us
|
| 52 |
+
GPU Operation Mode
|
| 53 |
+
Current : N/A
|
| 54 |
+
Pending : N/A
|
| 55 |
+
GPU C2C Mode : Disabled
|
| 56 |
+
GPU Virtualization Mode
|
| 57 |
+
Virtualization Mode : None
|
| 58 |
+
Host VGPU Mode : N/A
|
| 59 |
+
vGPU Heterogeneous Mode : N/A
|
| 60 |
+
GPU Recovery Action : None
|
| 61 |
+
GSP Firmware Version : 595.71.05
|
| 62 |
+
IBMNPU
|
| 63 |
+
Relaxed Ordering Mode : N/A
|
| 64 |
+
PCI
|
| 65 |
+
Bus : 0x17
|
| 66 |
+
Device : 0x00
|
| 67 |
+
Domain : 0x0000
|
| 68 |
+
Base Classcode : 0x3
|
| 69 |
+
Sub Classcode : 0x2
|
| 70 |
+
Device Id : 0x290110DE
|
| 71 |
+
Bus Id : 00000000:17:00.0
|
| 72 |
+
Sub System Id : 0x199910DE
|
| 73 |
+
GPU Link Info
|
| 74 |
+
PCIe Generation
|
| 75 |
+
Max : 5
|
| 76 |
+
Current : 5
|
| 77 |
+
Device Current : 5
|
| 78 |
+
Device Max : 5
|
| 79 |
+
Host Max : 5
|
| 80 |
+
Link Width
|
| 81 |
+
Max : 16x
|
| 82 |
+
Current : 16x
|
| 83 |
+
Bridge Chip
|
| 84 |
+
Type : N/A
|
| 85 |
+
Firmware : N/A
|
| 86 |
+
Replays Since Reset : 0
|
| 87 |
+
Replay Number Rollovers : 0
|
| 88 |
+
Tx Throughput : 814 KB/s
|
| 89 |
+
Rx Throughput : 553 KB/s
|
| 90 |
+
Atomic Caps Outbound : FETCHADD_32 FETCHADD_64 SWAP_32 SWAP_64 CAS_32 CAS_64
|
| 91 |
+
Atomic Caps Inbound : FETCHADD_32 FETCHADD_64 SWAP_32 SWAP_64 CAS_32 CAS_64
|
| 92 |
+
Fan Speed : N/A
|
| 93 |
+
Performance State : P0
|
| 94 |
+
Clocks Event Reasons
|
| 95 |
+
Idle : Not Active
|
| 96 |
+
Applications Clocks Setting : Not Active
|
| 97 |
+
SW Power Cap : Not Active
|
| 98 |
+
HW Slowdown : Not Active
|
| 99 |
+
HW Thermal Slowdown : Not Active
|
| 100 |
+
HW Power Brake Slowdown : Not Active
|
| 101 |
+
Sync Boost : Not Active
|
| 102 |
+
SW Thermal Slowdown : Not Active
|
| 103 |
+
Display Clock Setting : Not Active
|
| 104 |
+
Clocks Event Reasons Counters
|
| 105 |
+
SW Power Capping : 35245214504 us
|
| 106 |
+
Sync Boost : 0 us
|
| 107 |
+
SW Thermal Slowdown : 0 us
|
| 108 |
+
HW Thermal Slowdown : 0 us
|
| 109 |
+
HW Power Braking : 0 us
|
| 110 |
+
Sparse Operation Mode : N/A
|
| 111 |
+
FB Memory Usage
|
| 112 |
+
Total : 183359 MiB
|
| 113 |
+
Reserved : 735 MiB
|
| 114 |
+
Used : 0 MiB
|
| 115 |
+
Free : 182625 MiB
|
| 116 |
+
BAR1 Memory Usage
|
| 117 |
+
Total : 262144 MiB
|
| 118 |
+
Used : 0 MiB
|
| 119 |
+
Free : 262144 MiB
|
| 120 |
+
Conf Compute Protected Memory Usage
|
| 121 |
+
Total : 0 MiB
|
| 122 |
+
Used : 0 MiB
|
| 123 |
+
Free : 0 MiB
|
| 124 |
+
Compute Mode : Default
|
| 125 |
+
Utilization
|
| 126 |
+
GPU : 0 %
|
| 127 |
+
Memory : 0 %
|
| 128 |
+
Encoder : 0 %
|
| 129 |
+
Decoder : 0 %
|
| 130 |
+
JPEG : 0 %
|
| 131 |
+
OFA : 0 %
|
| 132 |
+
Encoder Stats
|
| 133 |
+
Active Sessions : 0
|
| 134 |
+
Average FPS : 0
|
| 135 |
+
Average Latency : 0
|
| 136 |
+
FBC Stats
|
| 137 |
+
Active Sessions : 0
|
| 138 |
+
Average FPS : 0
|
| 139 |
+
Average Latency : 0
|
| 140 |
+
DRAM Encryption Mode
|
| 141 |
+
Current : N/A
|
| 142 |
+
Pending : N/A
|
| 143 |
+
ECC Mode
|
| 144 |
+
Current : Enabled
|
| 145 |
+
Pending : Enabled
|
| 146 |
+
ECC Errors
|
| 147 |
+
Volatile
|
| 148 |
+
SRAM Correctable : 0
|
| 149 |
+
SRAM Uncorrectable Parity : 0
|
| 150 |
+
SRAM Uncorrectable SEC-DED : 0
|
| 151 |
+
DRAM Correctable : 0
|
| 152 |
+
DRAM Uncorrectable : 0
|
| 153 |
+
Aggregate
|
| 154 |
+
SRAM Correctable : 0
|
| 155 |
+
SRAM Uncorrectable Parity : 0
|
| 156 |
+
SRAM Uncorrectable SEC-DED : 0
|
| 157 |
+
DRAM Correctable : 0
|
| 158 |
+
DRAM Uncorrectable : 0
|
| 159 |
+
SRAM Threshold Exceeded : No
|
| 160 |
+
Aggregate Uncorrectable SRAM Sources
|
| 161 |
+
SRAM L2 : 0
|
| 162 |
+
SRAM SM : 0
|
| 163 |
+
SRAM Microcontroller : 0
|
| 164 |
+
SRAM PCIE : 0
|
| 165 |
+
SRAM Other : 0
|
| 166 |
+
Channel Repair Pending : No
|
| 167 |
+
TPC Repair Pending : No
|
| 168 |
+
Unrepairable Memory : No
|
| 169 |
+
Retired Pages
|
| 170 |
+
Single Bit ECC : N/A
|
| 171 |
+
Double Bit ECC : N/A
|
| 172 |
+
Pending Page Blacklist : N/A
|
| 173 |
+
Remapped Rows
|
| 174 |
+
Correctable Error : 0
|
| 175 |
+
Inactive Correctable Error : 0
|
| 176 |
+
Uncorrectable Error : 0
|
| 177 |
+
Inactive Uncorrectable Error : 0
|
| 178 |
+
Pending : No
|
| 179 |
+
Remapping Failure Occurred : No
|
| 180 |
+
Bank Remap Availability Histogram
|
| 181 |
+
Max : 3840 bank(s)
|
| 182 |
+
High : 0 bank(s)
|
| 183 |
+
Partial : 0 bank(s)
|
| 184 |
+
Low : 0 bank(s)
|
| 185 |
+
None : 0 bank(s)
|
| 186 |
+
Temperature
|
| 187 |
+
GPU Current Temp : 31 C
|
| 188 |
+
GPU T.Limit Temp : 57 C
|
| 189 |
+
GPU Shutdown T.Limit Temp : -5 C
|
| 190 |
+
GPU Slowdown T.Limit Temp : -3 C
|
| 191 |
+
GPU Max Operating T.Limit Temp : 0 C
|
| 192 |
+
GPU Target Temperature : N/A
|
| 193 |
+
Memory Current Temp : 31 C
|
| 194 |
+
Memory Max Operating T.Limit Temp : 0 C
|
| 195 |
+
GPU Power Readings
|
| 196 |
+
Average Power Draw : 248.19 W
|
| 197 |
+
Instantaneous Power Draw : 248.11 W
|
| 198 |
+
Current Power Limit : 1000.00 W
|
| 199 |
+
Requested Power Limit : 1000.00 W
|
| 200 |
+
Default Power Limit : 1000.00 W
|
| 201 |
+
Min Power Limit : 200.00 W
|
| 202 |
+
Max Power Limit : 1000.00 W
|
| 203 |
+
GPU Memory Power Readings
|
| 204 |
+
Average Power Draw : 24.90 W
|
| 205 |
+
Instantaneous Power Draw : N/A
|
| 206 |
+
Module Power Readings
|
| 207 |
+
Average Power Draw : N/A
|
| 208 |
+
Instantaneous Power Draw : N/A
|
| 209 |
+
Current Power Limit : N/A
|
| 210 |
+
Requested Power Limit : N/A
|
| 211 |
+
Default Power Limit : N/A
|
| 212 |
+
Min Power Limit : N/A
|
| 213 |
+
Max Power Limit : N/A
|
| 214 |
+
Power Smoothing : Insufficient Permissions
|
| 215 |
+
Workload Power Profiles
|
| 216 |
+
Requested Profiles : N/A
|
| 217 |
+
Enforced Profiles : N/A
|
| 218 |
+
EDPp Multiplier : 100.00%
|
| 219 |
+
Clocks
|
| 220 |
+
Graphics : 1965 MHz
|
| 221 |
+
SM : 1965 MHz
|
| 222 |
+
Memory : 3996 MHz
|
| 223 |
+
Video : 1860 MHz
|
| 224 |
+
Applications Clocks
|
| 225 |
+
Graphics : Requested functionality has been deprecated
|
| 226 |
+
Memory : Requested functionality has been deprecated
|
| 227 |
+
Default Applications Clocks
|
| 228 |
+
Graphics : Requested functionality has been deprecated
|
| 229 |
+
Memory : Requested functionality has been deprecated
|
| 230 |
+
Deferred Clocks
|
| 231 |
+
Memory : N/A
|
| 232 |
+
Max Clocks
|
| 233 |
+
Graphics : 1965 MHz
|
| 234 |
+
SM : 1965 MHz
|
| 235 |
+
Memory : 3996 MHz
|
| 236 |
+
Video : 1965 MHz
|
| 237 |
+
Max Customer Boost Clocks
|
| 238 |
+
Graphics : 1965 MHz
|
| 239 |
+
Clock Policy
|
| 240 |
+
Auto Boost : N/A
|
| 241 |
+
Auto Boost Default : N/A
|
| 242 |
+
Fabric
|
| 243 |
+
State : Completed
|
| 244 |
+
Status : Success
|
| 245 |
+
CliqueId : 0
|
| 246 |
+
ClusterUUID : 00000000-0000-0000-0000-000000000000
|
| 247 |
+
Health
|
| 248 |
+
Summary : Healthy
|
| 249 |
+
Bandwidth : N/A
|
| 250 |
+
Route Recovery in progress : N/A
|
| 251 |
+
Route Unhealthy : N/A
|
| 252 |
+
Access Timeout Recovery : False
|
| 253 |
+
Incorrect Configuration : N/A
|
| 254 |
+
Partition Assigned : N/A
|
| 255 |
+
Processes : None
|
| 256 |
+
Capabilities
|
| 257 |
+
EGM : disabled
|
| 258 |
+
|
benchmark/environment/pip-freeze.txt
ADDED
|
@@ -0,0 +1,170 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
accelerate==1.14.0
|
| 2 |
+
annotated-doc==0.0.4
|
| 3 |
+
anyio==4.11.0
|
| 4 |
+
argon2-cffi==25.1.0
|
| 5 |
+
argon2-cffi-bindings==25.1.0
|
| 6 |
+
arrow==1.3.0
|
| 7 |
+
asttokens==3.0.0
|
| 8 |
+
async-lru==2.0.5
|
| 9 |
+
attrs==25.4.0
|
| 10 |
+
babel==2.17.0
|
| 11 |
+
beautifulsoup4==4.14.2
|
| 12 |
+
bleach==6.2.0
|
| 13 |
+
blinker==1.7.0
|
| 14 |
+
certifi==2025.10.5
|
| 15 |
+
cffi==2.0.0
|
| 16 |
+
charset-normalizer==3.4.3
|
| 17 |
+
click==8.4.2
|
| 18 |
+
comm==0.2.3
|
| 19 |
+
cryptography==41.0.7
|
| 20 |
+
dbus-python==1.3.2
|
| 21 |
+
debugpy==1.8.17
|
| 22 |
+
decorator==5.2.1
|
| 23 |
+
decord==0.6.0
|
| 24 |
+
defusedxml==0.7.1
|
| 25 |
+
diffusers==0.39.0
|
| 26 |
+
distlib==0.4.0
|
| 27 |
+
distro==1.9.0
|
| 28 |
+
executing==2.2.1
|
| 29 |
+
fastjsonschema==2.21.2
|
| 30 |
+
filelock==3.20.0
|
| 31 |
+
fqdn==1.5.1
|
| 32 |
+
fsspec==2024.6.1
|
| 33 |
+
h11==0.16.0
|
| 34 |
+
hf-xet==1.5.1
|
| 35 |
+
hf_transfer==0.1.9
|
| 36 |
+
httpcore==1.0.9
|
| 37 |
+
httplib2==0.20.4
|
| 38 |
+
httpx==0.28.1
|
| 39 |
+
huggingface_hub==1.23.0
|
| 40 |
+
idna==3.10
|
| 41 |
+
ImageIO==2.37.3
|
| 42 |
+
imageio-ffmpeg==0.6.0
|
| 43 |
+
importlib_metadata==9.0.0
|
| 44 |
+
ipykernel==6.30.1
|
| 45 |
+
ipython==9.6.0
|
| 46 |
+
ipython_pygments_lexers==1.1.1
|
| 47 |
+
ipywidgets==8.1.7
|
| 48 |
+
isoduration==20.11.0
|
| 49 |
+
jedi==0.19.2
|
| 50 |
+
Jinja2==3.1.6
|
| 51 |
+
json5==0.12.1
|
| 52 |
+
json_repair==0.61.2
|
| 53 |
+
jsonpointer==3.0.0
|
| 54 |
+
jsonschema==4.25.1
|
| 55 |
+
jsonschema-specifications==2025.9.1
|
| 56 |
+
jupyter-archive==3.4.0
|
| 57 |
+
jupyter-events==0.12.0
|
| 58 |
+
jupyter-lsp==2.3.0
|
| 59 |
+
jupyter_client==8.6.3
|
| 60 |
+
jupyter_core==5.8.1
|
| 61 |
+
jupyter_server==2.17.0
|
| 62 |
+
jupyter_server_terminals==0.5.3
|
| 63 |
+
jupyterlab==4.4.9
|
| 64 |
+
jupyterlab_pygments==0.3.0
|
| 65 |
+
jupyterlab_server==2.27.3
|
| 66 |
+
jupyterlab_widgets==3.0.15
|
| 67 |
+
lark==1.3.0
|
| 68 |
+
launchpadlib==1.11.0
|
| 69 |
+
lazr.restfulclient==0.14.6
|
| 70 |
+
lazr.uri==1.0.6
|
| 71 |
+
lazy-loader==0.5
|
| 72 |
+
-e git+https://github.com/Robbyant/lingbot-video.git@a2bb04b78edd848500dc27a26e035a95442ae186#egg=lingbot_video
|
| 73 |
+
lpips==0.1.4
|
| 74 |
+
markdown-it-py==4.2.0
|
| 75 |
+
MarkupSafe==3.0.3
|
| 76 |
+
matplotlib-inline==0.1.7
|
| 77 |
+
mdurl==0.1.2
|
| 78 |
+
mistune==3.1.4
|
| 79 |
+
mpmath==1.3.0
|
| 80 |
+
nbclient==0.10.2
|
| 81 |
+
nbconvert==7.16.6
|
| 82 |
+
nbformat==5.10.4
|
| 83 |
+
nest-asyncio==1.6.0
|
| 84 |
+
networkx==3.3
|
| 85 |
+
notebook==7.4.2
|
| 86 |
+
notebook_shim==0.2.4
|
| 87 |
+
numpy==2.1.2
|
| 88 |
+
nvidia-cublas-cu12==12.8.4.1
|
| 89 |
+
nvidia-cuda-cupti-cu12==12.8.90
|
| 90 |
+
nvidia-cuda-nvrtc-cu12==12.8.93
|
| 91 |
+
nvidia-cuda-runtime-cu12==12.8.90
|
| 92 |
+
nvidia-cudnn-cu12==9.10.2.21
|
| 93 |
+
nvidia-cufft-cu12==11.3.3.83
|
| 94 |
+
nvidia-cufile-cu12==1.13.1.3
|
| 95 |
+
nvidia-curand-cu12==10.3.9.90
|
| 96 |
+
nvidia-cusolver-cu12==11.7.3.90
|
| 97 |
+
nvidia-cusparse-cu12==12.5.8.93
|
| 98 |
+
nvidia-cusparselt-cu12==0.7.1
|
| 99 |
+
nvidia-nccl-cu12==2.27.3
|
| 100 |
+
nvidia-nvjitlink-cu12==12.8.93
|
| 101 |
+
nvidia-nvtx-cu12==12.8.90
|
| 102 |
+
oauthlib==3.2.2
|
| 103 |
+
packaging==25.0
|
| 104 |
+
pandocfilters==1.5.1
|
| 105 |
+
parso==0.8.5
|
| 106 |
+
peft==0.19.1
|
| 107 |
+
pexpect==4.9.0
|
| 108 |
+
pillow==12.3.0
|
| 109 |
+
platformdirs==4.5.0
|
| 110 |
+
prometheus_client==0.23.1
|
| 111 |
+
prompt_toolkit==3.0.52
|
| 112 |
+
psutil==7.2.2
|
| 113 |
+
ptyprocess==0.7.0
|
| 114 |
+
pure_eval==0.2.3
|
| 115 |
+
pycparser==2.23
|
| 116 |
+
Pygments==2.19.2
|
| 117 |
+
PyGObject==3.48.2
|
| 118 |
+
PyJWT==2.7.0
|
| 119 |
+
pyparsing==3.1.1
|
| 120 |
+
python-apt==2.7.7+ubuntu5
|
| 121 |
+
python-dateutil==2.9.0.post0
|
| 122 |
+
python-json-logger==4.0.0
|
| 123 |
+
PyYAML==6.0.3
|
| 124 |
+
pyzmq==27.1.0
|
| 125 |
+
referencing==0.36.2
|
| 126 |
+
regex==2026.6.28
|
| 127 |
+
requests==2.34.2
|
| 128 |
+
rfc3339-validator==0.1.4
|
| 129 |
+
rfc3986-validator==0.1.1
|
| 130 |
+
rfc3987-syntax==1.1.0
|
| 131 |
+
rich==15.0.0
|
| 132 |
+
rpds-py==0.27.1
|
| 133 |
+
safetensors==0.8.0
|
| 134 |
+
scikit-image==0.26.0
|
| 135 |
+
scipy==1.18.0
|
| 136 |
+
-e git+https://github.com/Disty0/sdnq.git@d841c383ff7be38728d4df829e17af4f15d4fd66#egg=sdnq
|
| 137 |
+
Send2Trash==1.8.3
|
| 138 |
+
setuptools==83.0.0
|
| 139 |
+
shellingham==1.5.4
|
| 140 |
+
six==1.16.0
|
| 141 |
+
sniffio==1.3.1
|
| 142 |
+
soupsieve==2.8
|
| 143 |
+
stack-data==0.6.3
|
| 144 |
+
sympy==1.13.3
|
| 145 |
+
terminado==0.18.1
|
| 146 |
+
tifffile==2026.6.1
|
| 147 |
+
tinycss2==1.4.0
|
| 148 |
+
tokenizers==0.22.2
|
| 149 |
+
torch==2.8.0+cu128
|
| 150 |
+
torchaudio==2.8.0+cu128
|
| 151 |
+
torchvision==0.23.0+cu128
|
| 152 |
+
tornado==6.5.2
|
| 153 |
+
tqdm==4.68.4
|
| 154 |
+
traitlets==5.14.3
|
| 155 |
+
transformers==5.8.1
|
| 156 |
+
triton==3.4.0
|
| 157 |
+
typer==0.26.8
|
| 158 |
+
types-python-dateutil==2.9.0.20251008
|
| 159 |
+
typing_extensions==4.15.0
|
| 160 |
+
uri-template==1.3.0
|
| 161 |
+
urllib3==2.5.0
|
| 162 |
+
virtualenv==20.34.0
|
| 163 |
+
wadllib==1.3.6
|
| 164 |
+
wcwidth==0.2.14
|
| 165 |
+
webcolors==24.11.1
|
| 166 |
+
webencodings==0.5.1
|
| 167 |
+
websocket-client==1.9.0
|
| 168 |
+
wheel==0.47.0
|
| 169 |
+
widgetsnbextension==4.0.14
|
| 170 |
+
zipp==4.1.0
|
benchmark/original.metrics.csv
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
model_key,variant,prompt_id,seed,generation_seconds,seconds_per_frame,torch_peak_allocated_mb,torch_peak_reserved_mb,peak_gpu_memory_used_mb,peak_gpu_utilization_percent,peak_process_rss_bytes,peak_system_used_bytes,output_size_bytes,output_sha256
|
| 2 |
+
moe,original,01_arctic_research_drone,4201,140.9023737310199,1.9301695031646562,112469.38330078125,124524.0,125646.0,100.0,2700222464.0,80392216576.0,366044,7a548d09dd53755c1fdd7e759aae7cf92fd57d555bdc13f28023ede4ffe5f95e
|
| 3 |
+
moe,original,02_macro_pastry_glaze,4202,139.39877224492375,1.9095722225332021,112468.59130859375,124524.0,126274.0,100.0,3164041216.0,90717732864.0,396167,0073c4f94ed8fc0b9bfaa93966ac627b8a66f31efe9547a78bc6934ee2ecfd82
|
| 4 |
+
moe,original,03_rain_market_lanterns,4203,139.6812952979235,1.9134424013414177,112469.681640625,124524.0,125646.0,100.0,3218972672.0,91209596928.0,430040,753f2016262fbd6a3644b28ad4f6243e15e8b55ef9474616a9c7d636f474090b
|
| 5 |
+
moe,original,04_underwater_manta,4204,139.6351271129679,1.9128099604516149,112469.43212890625,124524.0,125646.0,100.0,3070435328.0,91173777408.0,359114,dcd3ffe1e7dd1f72f86e256d0db51acfb6661b880bfb0b2a5b798c9609cae24a
|
| 6 |
+
moe,original,05_factory_battery_cells,4205,139.8160271209199,1.9152880427523273,112470.375,124524.0,125646.0,100.0,3052527616.0,92903628800.0,324116,8a08377ecb3877e3ab261ca1840450dfa28a18ccc6cc59679002da1ff7750b49
|
benchmark/original.metrics.jsonl
ADDED
|
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"batch_cfg": false, "cold_generation": true, "ffprobe": {"programs": [], "streams": [{"codec_name": "h264", "duration": "3.041667", "height": 480, "nb_frames": "73", "pix_fmt": "yuv420p", "r_frame_rate": "24/1", "width": 832}]}, "fps": 24, "generation_seconds": 140.9023737310199, "guidance_scale": 3.0, "height": 480, "model_key": "moe", "negative_prompt": "{\"universal_negative\": {\"visual_quality\": [\"low quality\", \"worst quality\", \"blurry\", \"pixelated\", \"jpeg artifacts\", \"low resolution\", \"unstable color\", \"color flicker\", \"underexposed\", \"overexposed\", \"invisible subject\", \"subject hidden in darkness\"], \"artistic_style\": [\"painting\", \"illustration\", \"drawing\", \"cartoon\", \"3d render\", \"cgi\", \"sketch\", \"digital art\"], \"composition_and_content\": [\"text\", \"watermark\", \"signature\", \"logo\", \"subtitles\", \"pillarboxed\", \"side bars\", \"portrait image in landscape frame\"], \"temporal_and_motion_stability\": [\"flickering\", \"jittery\", \"motion blur\", \"temporal inconsistency\", \"warping\", \"morphing\", \"incoherent motion\", \"unnatural movement\", \"static object with sudden jump\", \"frame-to-frame inconsistency\"], \"material_and_structure\": [\"plastic-like glass\", \"unrealistic texture\", \"deformed bottle\", \"liquid freezing improperly\", \"distorted reflections\"]}}", "null_cond_clone_zero": false, "num_frames": 73, "num_inference_steps": 40, "nvidia_smi_after": {"memory.total": "183359", "memory.used": "125646", "name": "NVIDIA B200", "power.draw": "260.66", "temperature.gpu": "43", "timestamp": "2026/07/10 08:30:06.448", "utilization.gpu": "0"}, "output_file": "results/benchmark/moe/original/videos/01_arctic_research_drone.mp4", "output_sha256": "7a548d09dd53755c1fdd7e759aae7cf92fd57d555bdc13f28023ede4ffe5f95e", "output_size_bytes": 366044, "peak_disk_used_bytes": 188024430592.0, "peak_gpu_memory_used_mb": 125646.0, "peak_gpu_power_watts": 977.6, "peak_gpu_utilization_percent": 100.0, "peak_process_rss_bytes": 2700222464.0, "peak_system_used_bytes": 80392216576.0, "prompt": "{\"comprehensive_description\":{\"scene_content_description\":\"A cinematic dusk shot of an Arctic research outpost on a wind-polished ice shelf under green aurora ribbons. A compact orange rover with a roof-mounted lidar mast drives slowly between blue instrument crates and slender weather antennas. Fine snow dust trails behind the tires, and warm light glows from the station windows against the cold cyan landscape. The atmosphere is realistic, high-detail documentary science footage with crisp snow texture, vapor in the air, and physically plausible shadows.\",\"camera_movement_description\":\"The camera begins as a low, smooth drone tracking shot behind the rover, then arcs gently left to reveal the glowing station and aurora-filled sky while keeping the rover dominant in frame.\"},\"camera_info\":{\"color\":\"Cold cyan with warm orange accents\",\"frame_size\":\"Wide\",\"shot_type_angle\":\"Low angle tracking shot\",\"lens_size\":\"Wide lens\",\"composition\":\"Rover in lower center, research station in the background, aurora above\",\"lighting\":\"Soft twilight with window glow\",\"lighting_type\":\"Natural dusk and practical interior lights\"},\"world_knowledge\":[],\"prominent_elements\":[{\"name\":\"orange autonomous research rover\",\"description\":\"A rugged six-wheeled scientific rover with orange body panels, black tires, a lidar mast, and small blinking status lights.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.7s]\",\"action\":\"drives forward slowly across packed snow, tires compressing the surface\"},{\"timestamp\":\"[0.7s - 1.5s]\",\"action\":\"turns slightly left as the lidar mast rotates\"},{\"timestamp\":\"[1.5s - 2.0s]\",\"action\":\"continues toward the lit station while fine snow trails behind\"}],\"location\":\"lower center of the frame\",\"relative_size\":\"dominant\",\"shape_and_color\":\"low rectangular orange body with black wheels and grey sensor mast\",\"texture\":\"matte painted metal, rubber tires, frost on edges\",\"appearance_details\":\"roof lidar, small antennas, narrow headlights, compact cargo rack\",\"relationship\":\"the main moving subject, traveling toward the research outpost\",\"orientation\":\"moving away from the camera and slightly left\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"Arctic research station\",\"description\":\"A modular polar station made of connected white containers with warm yellow light in the windows.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"remains stationary as wind-blown snow passes in front of it\"}],\"location\":\"middle background\",\"relative_size\":\"large\",\"shape_and_color\":\"rectangular white modules with dark seams and glowing windows\",\"texture\":\"frosted metal panels and glass\",\"appearance_details\":\"small stairs, railings, antennas, blue equipment crates nearby\",\"relationship\":\"destination of the rover\",\"orientation\":\"angled three-quarter view\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"aurora ribbons\",\"description\":\"Green aurora curtains stretching across the darkening polar sky.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"shimmer slowly and drift laterally across the sky\"}],\"location\":\"upper half of the frame\",\"relative_size\":\"large\",\"shape_and_color\":\"soft green luminous ribbons\",\"texture\":\"translucent atmospheric glow\",\"appearance_details\":\"layered folds and faint stars behind\",\"relationship\":\"dominates the sky above the station\",\"orientation\":\"horizontal sweeping arcs\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":true,\"number_of_objects\":\"many luminous bands\"}]}", "prompt_file": "01_arctic_research_drone.json", "prompt_id": "01_arctic_research_drone", "raw_frames_file": "results/benchmark/moe/original/raw_frames/01_arctic_research_drone.npy", "raw_frames_sha256": "1c1cad878f2c8f39889f0acb71d6ebaf1b12222ff43ddcd8419e54e9cf00d590", "raw_frames_size_bytes": 87459968, "sample_count": 488, "sample_interval_seconds": 0.25, "seconds_per_frame": 1.9301695031646562, "seed": 4201, "shift": 3.0, "torch_allocated_after_mb": 103497.9833984375, "torch_peak_allocated_mb": 112469.38330078125, "torch_peak_reserved_mb": 124524.0, "torch_reserved_after_mb": 124524.0, "variant": "original", "width": 832}
|
| 2 |
+
{"batch_cfg": false, "cold_generation": false, "ffprobe": {"programs": [], "streams": [{"codec_name": "h264", "duration": "3.041667", "height": 480, "nb_frames": "73", "pix_fmt": "yuv420p", "r_frame_rate": "24/1", "width": 832}]}, "fps": 24, "generation_seconds": 139.39877224492375, "guidance_scale": 3.0, "height": 480, "model_key": "moe", "negative_prompt": "{\"universal_negative\": {\"visual_quality\": [\"low quality\", \"worst quality\", \"blurry\", \"pixelated\", \"jpeg artifacts\", \"low resolution\", \"unstable color\", \"color flicker\", \"underexposed\", \"overexposed\", \"invisible subject\", \"subject hidden in darkness\"], \"artistic_style\": [\"painting\", \"illustration\", \"drawing\", \"cartoon\", \"3d render\", \"cgi\", \"sketch\", \"digital art\"], \"composition_and_content\": [\"text\", \"watermark\", \"signature\", \"logo\", \"subtitles\", \"pillarboxed\", \"side bars\", \"portrait image in landscape frame\"], \"temporal_and_motion_stability\": [\"flickering\", \"jittery\", \"motion blur\", \"temporal inconsistency\", \"warping\", \"morphing\", \"incoherent motion\", \"unnatural movement\", \"static object with sudden jump\", \"frame-to-frame inconsistency\"], \"material_and_structure\": [\"plastic-like glass\", \"unrealistic texture\", \"deformed bottle\", \"liquid freezing improperly\", \"distorted reflections\"]}}", "null_cond_clone_zero": false, "num_frames": 73, "num_inference_steps": 40, "nvidia_smi_after": {"memory.total": "183359", "memory.used": "125646", "name": "NVIDIA B200", "power.draw": "260.70", "temperature.gpu": "43", "timestamp": "2026/07/10 08:32:29.979", "utilization.gpu": "0"}, "output_file": "results/benchmark/moe/original/videos/02_macro_pastry_glaze.mp4", "output_sha256": "0073c4f94ed8fc0b9bfaa93966ac627b8a66f31efe9547a78bc6934ee2ecfd82", "output_size_bytes": 396167, "peak_disk_used_bytes": 188112715776.0, "peak_gpu_memory_used_mb": 126274.0, "peak_gpu_power_watts": 969.78, "peak_gpu_utilization_percent": 100.0, "peak_process_rss_bytes": 3164041216.0, "peak_system_used_bytes": 90717732864.0, "prompt": "{\"comprehensive_description\":{\"scene_content_description\":\"An extreme macro food cinematography shot in a professional pastry kitchen. A glossy dark chocolate dome dessert sits on a rotating metal turntable. A thin stream of amber caramel glaze pours from above, flowing over the curved surface in slow ribbons and revealing tiny reflections from softbox lights. Crushed pistachio dust and edible gold flakes sit around the base on a matte black plate. The look is photorealistic, tactile, high-speed macro video with shallow depth of field and no text.\",\"camera_movement_description\":\"The camera holds a close macro angle with a subtle push-in while the dessert rotates clockwise on the turntable.\"},\"camera_info\":{\"color\":\"Deep brown, amber, green pistachio, gold highlights\",\"frame_size\":\"Extreme close up\",\"shot_type_angle\":\"Slight high angle\",\"lens_size\":\"Macro lens\",\"composition\":\"Dessert centered with glaze entering from top\",\"lighting\":\"Large softbox reflections\",\"lighting_type\":\"Controlled studio kitchen lighting\"},\"world_knowledge\":[],\"prominent_elements\":[{\"name\":\"chocolate dome dessert\",\"description\":\"A small mirror-glazed chocolate dome on a black plate, surrounded by pistachio dust and gold flakes.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"rotates slowly clockwise on a metal turntable\"}],\"location\":\"center of the frame\",\"relative_size\":\"dominant\",\"shape_and_color\":\"hemispherical dark chocolate dome on round black plate\",\"texture\":\"mirror-gloss surface with smooth curved reflections\",\"appearance_details\":\"tiny gold flakes, green pistachio crumbs, clean pastry plating\",\"relationship\":\"receives the caramel glaze\",\"orientation\":\"upright on the turntable\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"caramel glaze stream\",\"description\":\"A viscous amber caramel stream pouring from above and spreading over the dessert.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.8s]\",\"action\":\"falls as a narrow glossy stream onto the crown of the dome\"},{\"timestamp\":\"[0.8s - 1.6s]\",\"action\":\"forms slow ribbons that slide down the curved chocolate surface\"},{\"timestamp\":\"[1.6s - 2.0s]\",\"action\":\"collects at the base in a thin shiny ring\"}],\"location\":\"entering from top center and flowing over the dessert\",\"relative_size\":\"medium\",\"shape_and_color\":\"thin amber liquid ribbons\",\"texture\":\"viscous, glossy, translucent caramel\",\"appearance_details\":\"bright highlights and tiny bubbles\",\"relationship\":\"coats the chocolate dome\",\"orientation\":\"vertical stream becoming downward trails\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"}]}", "prompt_file": "02_macro_pastry_glaze.json", "prompt_id": "02_macro_pastry_glaze", "raw_frames_file": "results/benchmark/moe/original/raw_frames/02_macro_pastry_glaze.npy", "raw_frames_sha256": "c96f6b8795b970785ca2c50c383ab9e6055b18c05496bbe0873aa67dbea063ec", "raw_frames_size_bytes": 87459968, "sample_count": 463, "sample_interval_seconds": 0.25, "seconds_per_frame": 1.9095722225332021, "seed": 4202, "shift": 3.0, "torch_allocated_after_mb": 103497.9833984375, "torch_peak_allocated_mb": 112468.59130859375, "torch_peak_reserved_mb": 124524.0, "torch_reserved_after_mb": 124524.0, "variant": "original", "width": 832}
|
| 3 |
+
{"batch_cfg": false, "cold_generation": false, "ffprobe": {"programs": [], "streams": [{"codec_name": "h264", "duration": "3.041667", "height": 480, "nb_frames": "73", "pix_fmt": "yuv420p", "r_frame_rate": "24/1", "width": 832}]}, "fps": 24, "generation_seconds": 139.6812952979235, "guidance_scale": 3.0, "height": 480, "model_key": "moe", "negative_prompt": "{\"universal_negative\": {\"visual_quality\": [\"low quality\", \"worst quality\", \"blurry\", \"pixelated\", \"jpeg artifacts\", \"low resolution\", \"unstable color\", \"color flicker\", \"underexposed\", \"overexposed\", \"invisible subject\", \"subject hidden in darkness\"], \"artistic_style\": [\"painting\", \"illustration\", \"drawing\", \"cartoon\", \"3d render\", \"cgi\", \"sketch\", \"digital art\"], \"composition_and_content\": [\"text\", \"watermark\", \"signature\", \"logo\", \"subtitles\", \"pillarboxed\", \"side bars\", \"portrait image in landscape frame\"], \"temporal_and_motion_stability\": [\"flickering\", \"jittery\", \"motion blur\", \"temporal inconsistency\", \"warping\", \"morphing\", \"incoherent motion\", \"unnatural movement\", \"static object with sudden jump\", \"frame-to-frame inconsistency\"], \"material_and_structure\": [\"plastic-like glass\", \"unrealistic texture\", \"deformed bottle\", \"liquid freezing improperly\", \"distorted reflections\"]}}", "null_cond_clone_zero": false, "num_frames": 73, "num_inference_steps": 40, "nvidia_smi_after": {"memory.total": "183359", "memory.used": "125646", "name": "NVIDIA B200", "power.draw": "260.53", "temperature.gpu": "43", "timestamp": "2026/07/10 08:34:52.790", "utilization.gpu": "0"}, "output_file": "results/benchmark/moe/original/videos/03_rain_market_lanterns.mp4", "output_sha256": "753f2016262fbd6a3644b28ad4f6243e15e8b55ef9474616a9c7d636f474090b", "output_size_bytes": 430040, "peak_disk_used_bytes": 188200804352.0, "peak_gpu_memory_used_mb": 125646.0, "peak_gpu_power_watts": 972.57, "peak_gpu_utilization_percent": 100.0, "peak_process_rss_bytes": 3218972672.0, "peak_system_used_bytes": 91209596928.0, "prompt": "{\"comprehensive_description\":{\"scene_content_description\":\"A lively rainy night market in a narrow old-city street, filmed like naturalistic cinema. Red paper lanterns hang from wooden stalls, steam rises from bowls of soup, and wet cobblestones mirror orange and teal reflections. A vendor in a dark apron slides a steaming bowl across a counter to a traveler in a blue raincoat. Umbrellas pass in the foreground, partially occluding the scene for a moment. The mood is warm, humid, detailed, and grounded, with no readable signs or subtitles.\",\"camera_movement_description\":\"The camera performs a slow lateral dolly from left to right at counter height, with foreground umbrellas briefly crossing the lens.\"},\"camera_info\":{\"color\":\"Warm lantern orange, teal rain reflections, red accents\",\"frame_size\":\"Medium wide\",\"shot_type_angle\":\"Eye level\",\"lens_size\":\"Medium lens\",\"composition\":\"Vendor stall centered, traveler on right, passing umbrellas in foreground\",\"lighting\":\"Lantern light mixed with rain reflections\",\"lighting_type\":\"Night practical lighting\"},\"world_knowledge\":[],\"prominent_elements\":[{\"name\":\"food vendor\",\"description\":\"A middle-aged street vendor wearing a dark apron and working behind a steaming wooden food stall.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.8s]\",\"action\":\"lifts a white ceramic bowl from a steaming pot\"},{\"timestamp\":\"[0.8s - 1.6s]\",\"action\":\"slides the bowl carefully across the counter\"},{\"timestamp\":\"[1.6s - 2.0s]\",\"action\":\"wipes condensation from the counter with a cloth\"}],\"location\":\"center-left behind the stall counter\",\"relative_size\":\"large\",\"shape_and_color\":\"human figure in dark apron under warm lantern light\",\"texture\":\"wet fabric, skin highlights, wood grain around the stall\",\"appearance_details\":\"rolled sleeves, focused expression, steam around hands\",\"relationship\":\"serves food to the traveler\",\"orientation\":\"facing right toward the customer\",\"pose\":\"leaning slightly forward over the counter\",\"expression\":\"focused and calm\",\"clothing\":\"dark apron over a muted shirt\",\"gender\":\"\",\"skin_tone_and_texture\":\"natural skin with rain-lit highlights\"},{\"name\":\"traveler in blue raincoat\",\"description\":\"A traveler wearing a blue hooded raincoat, waiting at the stall with wet sleeves and a folded umbrella.\",\"actions\":[{\"timestamp\":\"[0.0s - 1.2s]\",\"action\":\"waits with hands near the counter, watching the bowl\"},{\"timestamp\":\"[1.2s - 2.0s]\",\"action\":\"reaches forward to receive the steaming bowl\"}],\"location\":\"right side of the frame\",\"relative_size\":\"medium\",\"shape_and_color\":\"blue hooded raincoat silhouette\",\"texture\":\"water beads on waterproof fabric\",\"appearance_details\":\"hood up, folded umbrella tucked under one arm\",\"relationship\":\"customer receiving food from the vendor\",\"orientation\":\"facing left toward the stall\",\"pose\":\"standing close to the counter\",\"expression\":\"expectant and tired\",\"clothing\":\"blue raincoat\",\"gender\":\"\",\"skin_tone_and_texture\":\"\"},{\"name\":\"rain and lantern reflections\",\"description\":\"Falling rain, steam, and glowing reflections on wet cobblestones.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"rain falls continuously while reflections ripple on the street\"}],\"location\":\"foreground and background throughout the frame\",\"relative_size\":\"large\",\"shape_and_color\":\"thin rain streaks, orange and teal reflected pools\",\"texture\":\"wet stone, vapor, glossy water\",\"appearance_details\":\"soft steam plumes, umbrellas passing close to lens\",\"relationship\":\"sets the atmosphere around the market\",\"orientation\":\"vertical rainfall and horizontal street reflections\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":true,\"number_of_objects\":\"many raindrops and reflections\"}]}", "prompt_file": "03_rain_market_lanterns.json", "prompt_id": "03_rain_market_lanterns", "raw_frames_file": "results/benchmark/moe/original/raw_frames/03_rain_market_lanterns.npy", "raw_frames_sha256": "8b0391e7276b04d2d802ce81b1f1a32d5e4cd0c976f55d2466daaa3a6c838a63", "raw_frames_size_bytes": 87459968, "sample_count": 486, "sample_interval_seconds": 0.25, "seconds_per_frame": 1.9134424013414177, "seed": 4203, "shift": 3.0, "torch_allocated_after_mb": 103497.9833984375, "torch_peak_allocated_mb": 112469.681640625, "torch_peak_reserved_mb": 124524.0, "torch_reserved_after_mb": 124524.0, "variant": "original", "width": 832}
|
| 4 |
+
{"batch_cfg": false, "cold_generation": false, "ffprobe": {"programs": [], "streams": [{"codec_name": "h264", "duration": "3.041667", "height": 480, "nb_frames": "73", "pix_fmt": "yuv420p", "r_frame_rate": "24/1", "width": 832}]}, "fps": 24, "generation_seconds": 139.6351271129679, "guidance_scale": 3.0, "height": 480, "model_key": "moe", "negative_prompt": "{\"universal_negative\": {\"visual_quality\": [\"low quality\", \"worst quality\", \"blurry\", \"pixelated\", \"jpeg artifacts\", \"low resolution\", \"unstable color\", \"color flicker\", \"underexposed\", \"overexposed\", \"invisible subject\", \"subject hidden in darkness\"], \"artistic_style\": [\"painting\", \"illustration\", \"drawing\", \"cartoon\", \"3d render\", \"cgi\", \"sketch\", \"digital art\"], \"composition_and_content\": [\"text\", \"watermark\", \"signature\", \"logo\", \"subtitles\", \"pillarboxed\", \"side bars\", \"portrait image in landscape frame\"], \"temporal_and_motion_stability\": [\"flickering\", \"jittery\", \"motion blur\", \"temporal inconsistency\", \"warping\", \"morphing\", \"incoherent motion\", \"unnatural movement\", \"static object with sudden jump\", \"frame-to-frame inconsistency\"], \"material_and_structure\": [\"plastic-like glass\", \"unrealistic texture\", \"deformed bottle\", \"liquid freezing improperly\", \"distorted reflections\"]}}", "null_cond_clone_zero": false, "num_frames": 73, "num_inference_steps": 40, "nvidia_smi_after": {"memory.total": "183359", "memory.used": "125646", "name": "NVIDIA B200", "power.draw": "261.37", "temperature.gpu": "43", "timestamp": "2026/07/10 08:37:15.363", "utilization.gpu": "0"}, "output_file": "results/benchmark/moe/original/videos/04_underwater_manta.mp4", "output_sha256": "dcd3ffe1e7dd1f72f86e256d0db51acfb6661b880bfb0b2a5b798c9609cae24a", "output_size_bytes": 359114, "peak_disk_used_bytes": 188289716224.0, "peak_gpu_memory_used_mb": 125646.0, "peak_gpu_power_watts": 978.59, "peak_gpu_utilization_percent": 100.0, "peak_process_rss_bytes": 3070435328.0, "peak_system_used_bytes": 91173777408.0, "prompt": "{\"comprehensive_description\":{\"scene_content_description\":\"A tranquil underwater wildlife scene in clear tropical water. A large manta ray glides above a colorful coral garden while a scuba diver hovers several meters behind, holding a small camera rig. Sunbeams ripple through the surface, schools of tiny silver fish split and rejoin around the manta, and suspended particles drift slowly. The image should feel like realistic underwater documentary footage with natural motion, soft blue-green color, and no fantasy elements.\",\"camera_movement_description\":\"The camera tracks smoothly beside the manta ray from a slightly lower angle, drifting forward with the current.\"},\"camera_info\":{\"color\":\"Blue-green water with coral reds and yellows\",\"frame_size\":\"Wide\",\"shot_type_angle\":\"Slight low angle\",\"lens_size\":\"Wide underwater lens\",\"composition\":\"Manta ray crossing the upper center, coral below, diver in background\",\"lighting\":\"Dappled sunlight\",\"lighting_type\":\"Natural underwater daylight\"},\"world_knowledge\":[],\"prominent_elements\":[{\"name\":\"manta ray\",\"description\":\"A large manta ray with broad triangular fins, dark top surface, and pale underside.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.8s]\",\"action\":\"glides from left to right with slow wing-like fin movement\"},{\"timestamp\":\"[0.8s - 1.5s]\",\"action\":\"tilts slightly upward as sunbeams cross its back\"},{\"timestamp\":\"[1.5s - 2.0s]\",\"action\":\"continues forward while small fish scatter around it\"}],\"location\":\"upper center moving toward right\",\"relative_size\":\"dominant\",\"shape_and_color\":\"wide diamond-like silhouette, dark grey top, pale underside\",\"texture\":\"smooth skin with subtle mottling\",\"appearance_details\":\"cephalic fins near mouth, long tail trailing behind\",\"relationship\":\"main wildlife subject of the video\",\"orientation\":\"moving left to right and slightly upward\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"scuba diver\",\"description\":\"A diver in black wetsuit and fins, carrying a small underwater camera rig.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"hovers calmly in the background, exhaling small bubble streams\"}],\"location\":\"mid-background behind the manta ray\",\"relative_size\":\"small\",\"shape_and_color\":\"human figure in black wetsuit with silver air tank\",\"texture\":\"neoprene suit, metal tank, glass mask\",\"appearance_details\":\"fins, mask, regulator, compact camera lights\",\"relationship\":\"observes the manta ray without touching it\",\"orientation\":\"facing the manta ray\",\"pose\":\"horizontal hover\",\"expression\":\"\",\"clothing\":\"black scuba gear\",\"gender\":\"\",\"skin_tone_and_texture\":\"\"},{\"name\":\"coral reef and fish\",\"description\":\"A colorful coral garden with small silver fish schooling above it.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"fish swirl and split around the manta while coral remains fixed\"}],\"location\":\"lower half of the frame\",\"relative_size\":\"large\",\"shape_and_color\":\"branching corals in red, yellow, and muted purple; silver fish clusters\",\"texture\":\"rough coral, shimmering fish scales\",\"appearance_details\":\"sand patches, sea fans, drifting particles\",\"relationship\":\"environment below the manta ray\",\"orientation\":\"reef spreads horizontally across the bottom\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":true,\"number_of_objects\":\"many fish and coral structures\"}]}", "prompt_file": "04_underwater_manta.json", "prompt_id": "04_underwater_manta", "raw_frames_file": "results/benchmark/moe/original/raw_frames/04_underwater_manta.npy", "raw_frames_sha256": "68c6be000f9405a55cd6625cef632beb4484dcf5d1ce3157448c4093c0340ee3", "raw_frames_size_bytes": 87459968, "sample_count": 479, "sample_interval_seconds": 0.25, "seconds_per_frame": 1.9128099604516149, "seed": 4204, "shift": 3.0, "torch_allocated_after_mb": 103498.359375, "torch_peak_allocated_mb": 112469.43212890625, "torch_peak_reserved_mb": 124524.0, "torch_reserved_after_mb": 124524.0, "variant": "original", "width": 832}
|
| 5 |
+
{"batch_cfg": false, "cold_generation": false, "ffprobe": {"programs": [], "streams": [{"codec_name": "h264", "duration": "3.041667", "height": 480, "nb_frames": "73", "pix_fmt": "yuv420p", "r_frame_rate": "24/1", "width": 832}]}, "fps": 24, "generation_seconds": 139.8160271209199, "guidance_scale": 3.0, "height": 480, "model_key": "moe", "negative_prompt": "{\"universal_negative\": {\"visual_quality\": [\"low quality\", \"worst quality\", \"blurry\", \"pixelated\", \"jpeg artifacts\", \"low resolution\", \"unstable color\", \"color flicker\", \"underexposed\", \"overexposed\", \"invisible subject\", \"subject hidden in darkness\"], \"artistic_style\": [\"painting\", \"illustration\", \"drawing\", \"cartoon\", \"3d render\", \"cgi\", \"sketch\", \"digital art\"], \"composition_and_content\": [\"text\", \"watermark\", \"signature\", \"logo\", \"subtitles\", \"pillarboxed\", \"side bars\", \"portrait image in landscape frame\"], \"temporal_and_motion_stability\": [\"flickering\", \"jittery\", \"motion blur\", \"temporal inconsistency\", \"warping\", \"morphing\", \"incoherent motion\", \"unnatural movement\", \"static object with sudden jump\", \"frame-to-frame inconsistency\"], \"material_and_structure\": [\"plastic-like glass\", \"unrealistic texture\", \"deformed bottle\", \"liquid freezing improperly\", \"distorted reflections\"]}}", "null_cond_clone_zero": false, "num_frames": 73, "num_inference_steps": 40, "nvidia_smi_after": {"memory.total": "183359", "memory.used": "125646", "name": "NVIDIA B200", "power.draw": "260.91", "temperature.gpu": "43", "timestamp": "2026/07/10 08:39:38.091", "utilization.gpu": "0"}, "output_file": "results/benchmark/moe/original/videos/05_factory_battery_cells.mp4", "output_sha256": "8a08377ecb3877e3ab261ca1840450dfa28a18ccc6cc59679002da1ff7750b49", "output_size_bytes": 324116, "peak_disk_used_bytes": 188377776128.0, "peak_gpu_memory_used_mb": 125646.0, "peak_gpu_power_watts": 968.58, "peak_gpu_utilization_percent": 100.0, "peak_process_rss_bytes": 3052527616.0, "peak_system_used_bytes": 92903628800.0, "prompt": "{\"comprehensive_description\":{\"scene_content_description\":\"A clean high-tech factory line assembling transparent solid-state battery cells. Two white robotic arms move with precise synchronized motion over a brushed steel conveyor. One arm lowers a translucent rectangular cell into a copper test fixture while the other arm scans it with a blue light bar. Tiny reflections slide across glass safety panels, and status lights pulse softly. The scene should look like realistic industrial automation footage, not CGI, with crisp metal surfaces and controlled motion.\",\"camera_movement_description\":\"The camera is locked off in a medium-wide three-quarter view with a subtle mechanical vibration, emphasizing repeatable robotic motion.\"},\"camera_info\":{\"color\":\"White, brushed steel, copper, blue inspection light\",\"frame_size\":\"Medium wide\",\"shot_type_angle\":\"Three-quarter eye-level industrial view\",\"lens_size\":\"Medium lens\",\"composition\":\"Two robotic arms framing the battery cell at center\",\"lighting\":\"Even overhead factory lighting with blue scanner highlight\",\"lighting_type\":\"Artificial industrial lighting\"},\"world_knowledge\":[],\"prominent_elements\":[{\"name\":\"left robotic arm\",\"description\":\"A white six-axis robotic arm with a vacuum gripper holding a transparent rectangular battery cell.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.7s]\",\"action\":\"moves downward from upper left with the transparent cell held steady\"},{\"timestamp\":\"[0.7s - 1.4s]\",\"action\":\"places the cell into a copper test fixture\"},{\"timestamp\":\"[1.4s - 2.0s]\",\"action\":\"releases the cell and retracts slightly upward\"}],\"location\":\"left side moving toward center\",\"relative_size\":\"large\",\"shape_and_color\":\"white articulated segments with black joints\",\"texture\":\"smooth painted metal and rubber vacuum cups\",\"appearance_details\":\"visible cable routing, small green status LED\",\"relationship\":\"positions the battery cell for testing\",\"orientation\":\"angled downward toward the center fixture\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"right robotic scanner arm\",\"description\":\"A second white robotic arm carrying a rectangular blue inspection light bar.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.9s]\",\"action\":\"waits above the copper fixture with blue scanner light dim\"},{\"timestamp\":\"[0.9s - 1.6s]\",\"action\":\"sweeps the blue light bar across the transparent cell\"},{\"timestamp\":\"[1.6s - 2.0s]\",\"action\":\"pauses as the scanner light pulses once\"}],\"location\":\"right side above the conveyor\",\"relative_size\":\"large\",\"shape_and_color\":\"white arm with glowing blue rectangular scanner\",\"texture\":\"smooth metal casing and glass scanner cover\",\"appearance_details\":\"blue light strip, black joints, compact sensor module\",\"relationship\":\"inspects the cell after placement\",\"orientation\":\"angled left toward the battery cell\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"transparent battery cell and copper fixture\",\"description\":\"A clear rectangular solid-state cell seated into a copper test fixture on a steel conveyor.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.7s]\",\"action\":\"moves with the left arm above the fixture\"},{\"timestamp\":\"[0.7s - 1.4s]\",\"action\":\"is lowered precisely into the fixture\"},{\"timestamp\":\"[1.4s - 2.0s]\",\"action\":\"remains stationary while blue inspection light passes over it\"}],\"location\":\"center of the frame on the conveyor\",\"relative_size\":\"medium\",\"shape_and_color\":\"transparent rectangle with faint internal layers, copper fixture\",\"texture\":\"clear glass-like cell, brushed copper, polished steel\",\"appearance_details\":\"thin internal laminate layers, small alignment pins, clean conveyor surface\",\"relationship\":\"object being assembled and inspected\",\"orientation\":\"horizontal in the fixture\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"}]}", "prompt_file": "05_factory_battery_cells.json", "prompt_id": "05_factory_battery_cells", "raw_frames_file": "results/benchmark/moe/original/raw_frames/05_factory_battery_cells.npy", "raw_frames_sha256": "6107a35b979814f559baa8670e63c84f02b13c6f77c111203adb8ea199766c61", "raw_frames_size_bytes": 87459968, "sample_count": 473, "sample_interval_seconds": 0.25, "seconds_per_frame": 1.9152880427523273, "seed": 4205, "shift": 3.0, "torch_allocated_after_mb": 103497.9833984375, "torch_peak_allocated_mb": 112470.375, "torch_peak_reserved_mb": 124524.0, "torch_reserved_after_mb": 124524.0, "variant": "original", "width": 832}
|
benchmark/paired.metrics.csv
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
prompt_id,seed,height,width,num_frames,fps,num_inference_steps,guidance_scale,shift,original_generation_seconds,original_seconds_per_frame,original_torch_peak_allocated_mb,original_torch_peak_reserved_mb,original_peak_gpu_memory_used_mb,original_peak_gpu_utilization_percent,original_peak_process_rss_bytes,original_peak_system_used_bytes,original_output_size_bytes,original_output_sha256,sdnq_generation_seconds,sdnq_seconds_per_frame,sdnq_torch_peak_allocated_mb,sdnq_torch_peak_reserved_mb,sdnq_peak_gpu_memory_used_mb,sdnq_peak_gpu_utilization_percent,sdnq_peak_process_rss_bytes,sdnq_peak_system_used_bytes,sdnq_output_size_bytes,sdnq_output_sha256
|
| 2 |
+
01_arctic_research_drone,4201,480,832,73,24,40,3.0,3.0,140.9023737310199,1.9301695031646562,112469.38330078125,124524.0,125646.0,100.0,2700222464.0,80392216576.0,366044,7a548d09dd53755c1fdd7e759aae7cf92fd57d555bdc13f28023ede4ffe5f95e,159.77318301890045,2.188673739984938,36067.29833984375,40756.0,41878.0,100.0,2920222720.0,138935304192.0,339957,d040042bdb79aacdcafe80a2c4bd621271e3056823a84621c25ee8fd4fff594b
|
| 3 |
+
02_macro_pastry_glaze,4202,480,832,73,24,40,3.0,3.0,139.39877224492375,1.9095722225332021,112468.59130859375,124524.0,126274.0,100.0,3164041216.0,90717732864.0,396167,0073c4f94ed8fc0b9bfaa93966ac627b8a66f31efe9547a78bc6934ee2ecfd82,158.2769890979398,2.1681779328484905,36065.68798828125,40756.0,42504.0,100.0,3105587200.0,90820800512.0,367773,bcb2caabb9e29427019ef3dcc7790a89910c1b570e65f85bac3f8a95dc1912db
|
| 4 |
+
03_rain_market_lanterns,4203,480,832,73,24,40,3.0,3.0,139.6812952979235,1.9134424013414177,112469.681640625,124524.0,125646.0,100.0,3218972672.0,91209596928.0,430040,753f2016262fbd6a3644b28ad4f6243e15e8b55ef9474616a9c7d636f474090b,158.72098272107542,2.1742600372750056,36067.5908203125,40756.0,41878.0,100.0,3175432192.0,91215613952.0,360438,e91a616e88278e062635d12bd204679caf879dafc107cb066cf4155c3efe2486
|
| 5 |
+
04_underwater_manta,4204,480,832,73,24,40,3.0,3.0,139.6351271129679,1.9128099604516149,112469.43212890625,124524.0,125646.0,100.0,3070435328.0,91173777408.0,359114,dcd3ffe1e7dd1f72f86e256d0db51acfb6661b880bfb0b2a5b798c9609cae24a,158.60668343293946,2.1726942936019102,36067.27880859375,40756.0,41878.0,100.0,3106979840.0,90012438528.0,359678,add7b1153f552850a2b095afa51998c4e3af4719f3d5c64425105f1af8c7458e
|
| 6 |
+
05_factory_battery_cells,4205,480,832,73,24,40,3.0,3.0,139.8160271209199,1.9152880427523273,112470.375,124524.0,125646.0,100.0,3052527616.0,92903628800.0,324116,8a08377ecb3877e3ab261ca1840450dfa28a18ccc6cc59679002da1ff7750b49,158.87487718299963,2.176368180589036,36067.6962890625,40756.0,41878.0,100.0,3307651072.0,87309594624.0,138406,7b6909a2aa49397f574338560cce4a3e47bb1930e55304f897523fa436fe7333
|
benchmark/paired.metrics.jsonl
ADDED
|
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"fps": 24, "guidance_scale": 3.0, "height": 480, "num_frames": 73, "num_inference_steps": 40, "original_generation_seconds": 140.9023737310199, "original_output_sha256": "7a548d09dd53755c1fdd7e759aae7cf92fd57d555bdc13f28023ede4ffe5f95e", "original_output_size_bytes": 366044, "original_peak_gpu_memory_used_mb": 125646.0, "original_peak_gpu_utilization_percent": 100.0, "original_peak_process_rss_bytes": 2700222464.0, "original_peak_system_used_bytes": 80392216576.0, "original_seconds_per_frame": 1.9301695031646562, "original_torch_peak_allocated_mb": 112469.38330078125, "original_torch_peak_reserved_mb": 124524.0, "prompt_id": "01_arctic_research_drone", "sdnq_generation_seconds": 159.77318301890045, "sdnq_output_sha256": "d040042bdb79aacdcafe80a2c4bd621271e3056823a84621c25ee8fd4fff594b", "sdnq_output_size_bytes": 339957, "sdnq_peak_gpu_memory_used_mb": 41878.0, "sdnq_peak_gpu_utilization_percent": 100.0, "sdnq_peak_process_rss_bytes": 2920222720.0, "sdnq_peak_system_used_bytes": 138935304192.0, "sdnq_seconds_per_frame": 2.188673739984938, "sdnq_torch_peak_allocated_mb": 36067.29833984375, "sdnq_torch_peak_reserved_mb": 40756.0, "seed": 4201, "shift": 3.0, "width": 832}
|
| 2 |
+
{"fps": 24, "guidance_scale": 3.0, "height": 480, "num_frames": 73, "num_inference_steps": 40, "original_generation_seconds": 139.39877224492375, "original_output_sha256": "0073c4f94ed8fc0b9bfaa93966ac627b8a66f31efe9547a78bc6934ee2ecfd82", "original_output_size_bytes": 396167, "original_peak_gpu_memory_used_mb": 126274.0, "original_peak_gpu_utilization_percent": 100.0, "original_peak_process_rss_bytes": 3164041216.0, "original_peak_system_used_bytes": 90717732864.0, "original_seconds_per_frame": 1.9095722225332021, "original_torch_peak_allocated_mb": 112468.59130859375, "original_torch_peak_reserved_mb": 124524.0, "prompt_id": "02_macro_pastry_glaze", "sdnq_generation_seconds": 158.2769890979398, "sdnq_output_sha256": "bcb2caabb9e29427019ef3dcc7790a89910c1b570e65f85bac3f8a95dc1912db", "sdnq_output_size_bytes": 367773, "sdnq_peak_gpu_memory_used_mb": 42504.0, "sdnq_peak_gpu_utilization_percent": 100.0, "sdnq_peak_process_rss_bytes": 3105587200.0, "sdnq_peak_system_used_bytes": 90820800512.0, "sdnq_seconds_per_frame": 2.1681779328484905, "sdnq_torch_peak_allocated_mb": 36065.68798828125, "sdnq_torch_peak_reserved_mb": 40756.0, "seed": 4202, "shift": 3.0, "width": 832}
|
| 3 |
+
{"fps": 24, "guidance_scale": 3.0, "height": 480, "num_frames": 73, "num_inference_steps": 40, "original_generation_seconds": 139.6812952979235, "original_output_sha256": "753f2016262fbd6a3644b28ad4f6243e15e8b55ef9474616a9c7d636f474090b", "original_output_size_bytes": 430040, "original_peak_gpu_memory_used_mb": 125646.0, "original_peak_gpu_utilization_percent": 100.0, "original_peak_process_rss_bytes": 3218972672.0, "original_peak_system_used_bytes": 91209596928.0, "original_seconds_per_frame": 1.9134424013414177, "original_torch_peak_allocated_mb": 112469.681640625, "original_torch_peak_reserved_mb": 124524.0, "prompt_id": "03_rain_market_lanterns", "sdnq_generation_seconds": 158.72098272107542, "sdnq_output_sha256": "e91a616e88278e062635d12bd204679caf879dafc107cb066cf4155c3efe2486", "sdnq_output_size_bytes": 360438, "sdnq_peak_gpu_memory_used_mb": 41878.0, "sdnq_peak_gpu_utilization_percent": 100.0, "sdnq_peak_process_rss_bytes": 3175432192.0, "sdnq_peak_system_used_bytes": 91215613952.0, "sdnq_seconds_per_frame": 2.1742600372750056, "sdnq_torch_peak_allocated_mb": 36067.5908203125, "sdnq_torch_peak_reserved_mb": 40756.0, "seed": 4203, "shift": 3.0, "width": 832}
|
| 4 |
+
{"fps": 24, "guidance_scale": 3.0, "height": 480, "num_frames": 73, "num_inference_steps": 40, "original_generation_seconds": 139.6351271129679, "original_output_sha256": "dcd3ffe1e7dd1f72f86e256d0db51acfb6661b880bfb0b2a5b798c9609cae24a", "original_output_size_bytes": 359114, "original_peak_gpu_memory_used_mb": 125646.0, "original_peak_gpu_utilization_percent": 100.0, "original_peak_process_rss_bytes": 3070435328.0, "original_peak_system_used_bytes": 91173777408.0, "original_seconds_per_frame": 1.9128099604516149, "original_torch_peak_allocated_mb": 112469.43212890625, "original_torch_peak_reserved_mb": 124524.0, "prompt_id": "04_underwater_manta", "sdnq_generation_seconds": 158.60668343293946, "sdnq_output_sha256": "add7b1153f552850a2b095afa51998c4e3af4719f3d5c64425105f1af8c7458e", "sdnq_output_size_bytes": 359678, "sdnq_peak_gpu_memory_used_mb": 41878.0, "sdnq_peak_gpu_utilization_percent": 100.0, "sdnq_peak_process_rss_bytes": 3106979840.0, "sdnq_peak_system_used_bytes": 90012438528.0, "sdnq_seconds_per_frame": 2.1726942936019102, "sdnq_torch_peak_allocated_mb": 36067.27880859375, "sdnq_torch_peak_reserved_mb": 40756.0, "seed": 4204, "shift": 3.0, "width": 832}
|
| 5 |
+
{"fps": 24, "guidance_scale": 3.0, "height": 480, "num_frames": 73, "num_inference_steps": 40, "original_generation_seconds": 139.8160271209199, "original_output_sha256": "8a08377ecb3877e3ab261ca1840450dfa28a18ccc6cc59679002da1ff7750b49", "original_output_size_bytes": 324116, "original_peak_gpu_memory_used_mb": 125646.0, "original_peak_gpu_utilization_percent": 100.0, "original_peak_process_rss_bytes": 3052527616.0, "original_peak_system_used_bytes": 92903628800.0, "original_seconds_per_frame": 1.9152880427523273, "original_torch_peak_allocated_mb": 112470.375, "original_torch_peak_reserved_mb": 124524.0, "prompt_id": "05_factory_battery_cells", "sdnq_generation_seconds": 158.87487718299963, "sdnq_output_sha256": "7b6909a2aa49397f574338560cce4a3e47bb1930e55304f897523fa436fe7333", "sdnq_output_size_bytes": 138406, "sdnq_peak_gpu_memory_used_mb": 41878.0, "sdnq_peak_gpu_utilization_percent": 100.0, "sdnq_peak_process_rss_bytes": 3307651072.0, "sdnq_peak_system_used_bytes": 87309594624.0, "sdnq_seconds_per_frame": 2.176368180589036, "sdnq_torch_peak_allocated_mb": 36067.6962890625, "sdnq_torch_peak_reserved_mb": 40756.0, "seed": 4205, "shift": 3.0, "width": 832}
|
benchmark/prompts/01_arctic_research_drone.json
ADDED
|
@@ -0,0 +1,95 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"caption": {
|
| 3 |
+
"comprehensive_description": {
|
| 4 |
+
"scene_content_description": "A cinematic dusk shot of an Arctic research outpost on a wind-polished ice shelf under green aurora ribbons. A compact orange rover with a roof-mounted lidar mast drives slowly between blue instrument crates and slender weather antennas. Fine snow dust trails behind the tires, and warm light glows from the station windows against the cold cyan landscape. The atmosphere is realistic, high-detail documentary science footage with crisp snow texture, vapor in the air, and physically plausible shadows.",
|
| 5 |
+
"camera_movement_description": "The camera begins as a low, smooth drone tracking shot behind the rover, then arcs gently left to reveal the glowing station and aurora-filled sky while keeping the rover dominant in frame."
|
| 6 |
+
},
|
| 7 |
+
"camera_info": {
|
| 8 |
+
"color": "Cold cyan with warm orange accents",
|
| 9 |
+
"frame_size": "Wide",
|
| 10 |
+
"shot_type_angle": "Low angle tracking shot",
|
| 11 |
+
"lens_size": "Wide lens",
|
| 12 |
+
"composition": "Rover in lower center, research station in the background, aurora above",
|
| 13 |
+
"lighting": "Soft twilight with window glow",
|
| 14 |
+
"lighting_type": "Natural dusk and practical interior lights"
|
| 15 |
+
},
|
| 16 |
+
"world_knowledge": [],
|
| 17 |
+
"prominent_elements": [
|
| 18 |
+
{
|
| 19 |
+
"name": "orange autonomous research rover",
|
| 20 |
+
"description": "A rugged six-wheeled scientific rover with orange body panels, black tires, a lidar mast, and small blinking status lights.",
|
| 21 |
+
"actions": [
|
| 22 |
+
{
|
| 23 |
+
"timestamp": "[0.0s - 0.7s]",
|
| 24 |
+
"action": "drives forward slowly across packed snow, tires compressing the surface"
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"timestamp": "[0.7s - 1.5s]",
|
| 28 |
+
"action": "turns slightly left as the lidar mast rotates"
|
| 29 |
+
},
|
| 30 |
+
{
|
| 31 |
+
"timestamp": "[1.5s - 2.0s]",
|
| 32 |
+
"action": "continues toward the lit station while fine snow trails behind"
|
| 33 |
+
}
|
| 34 |
+
],
|
| 35 |
+
"location": "lower center of the frame",
|
| 36 |
+
"relative_size": "dominant",
|
| 37 |
+
"shape_and_color": "low rectangular orange body with black wheels and grey sensor mast",
|
| 38 |
+
"texture": "matte painted metal, rubber tires, frost on edges",
|
| 39 |
+
"appearance_details": "roof lidar, small antennas, narrow headlights, compact cargo rack",
|
| 40 |
+
"relationship": "the main moving subject, traveling toward the research outpost",
|
| 41 |
+
"orientation": "moving away from the camera and slightly left",
|
| 42 |
+
"pose": "",
|
| 43 |
+
"expression": "",
|
| 44 |
+
"clothing": "",
|
| 45 |
+
"is_cluster": false,
|
| 46 |
+
"number_of_objects": ""
|
| 47 |
+
},
|
| 48 |
+
{
|
| 49 |
+
"name": "Arctic research station",
|
| 50 |
+
"description": "A modular polar station made of connected white containers with warm yellow light in the windows.",
|
| 51 |
+
"actions": [
|
| 52 |
+
{
|
| 53 |
+
"timestamp": "[0.0s - 2.0s]",
|
| 54 |
+
"action": "remains stationary as wind-blown snow passes in front of it"
|
| 55 |
+
}
|
| 56 |
+
],
|
| 57 |
+
"location": "middle background",
|
| 58 |
+
"relative_size": "large",
|
| 59 |
+
"shape_and_color": "rectangular white modules with dark seams and glowing windows",
|
| 60 |
+
"texture": "frosted metal panels and glass",
|
| 61 |
+
"appearance_details": "small stairs, railings, antennas, blue equipment crates nearby",
|
| 62 |
+
"relationship": "destination of the rover",
|
| 63 |
+
"orientation": "angled three-quarter view",
|
| 64 |
+
"pose": "",
|
| 65 |
+
"expression": "",
|
| 66 |
+
"clothing": "",
|
| 67 |
+
"is_cluster": false,
|
| 68 |
+
"number_of_objects": ""
|
| 69 |
+
},
|
| 70 |
+
{
|
| 71 |
+
"name": "aurora ribbons",
|
| 72 |
+
"description": "Green aurora curtains stretching across the darkening polar sky.",
|
| 73 |
+
"actions": [
|
| 74 |
+
{
|
| 75 |
+
"timestamp": "[0.0s - 2.0s]",
|
| 76 |
+
"action": "shimmer slowly and drift laterally across the sky"
|
| 77 |
+
}
|
| 78 |
+
],
|
| 79 |
+
"location": "upper half of the frame",
|
| 80 |
+
"relative_size": "large",
|
| 81 |
+
"shape_and_color": "soft green luminous ribbons",
|
| 82 |
+
"texture": "translucent atmospheric glow",
|
| 83 |
+
"appearance_details": "layered folds and faint stars behind",
|
| 84 |
+
"relationship": "dominates the sky above the station",
|
| 85 |
+
"orientation": "horizontal sweeping arcs",
|
| 86 |
+
"pose": "",
|
| 87 |
+
"expression": "",
|
| 88 |
+
"clothing": "",
|
| 89 |
+
"is_cluster": true,
|
| 90 |
+
"number_of_objects": "many luminous bands"
|
| 91 |
+
}
|
| 92 |
+
]
|
| 93 |
+
},
|
| 94 |
+
"duration": 2.0
|
| 95 |
+
}
|
benchmark/prompts/02_macro_pastry_glaze.json
ADDED
|
@@ -0,0 +1,73 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"caption": {
|
| 3 |
+
"comprehensive_description": {
|
| 4 |
+
"scene_content_description": "An extreme macro food cinematography shot in a professional pastry kitchen. A glossy dark chocolate dome dessert sits on a rotating metal turntable. A thin stream of amber caramel glaze pours from above, flowing over the curved surface in slow ribbons and revealing tiny reflections from softbox lights. Crushed pistachio dust and edible gold flakes sit around the base on a matte black plate. The look is photorealistic, tactile, high-speed macro video with shallow depth of field and no text.",
|
| 5 |
+
"camera_movement_description": "The camera holds a close macro angle with a subtle push-in while the dessert rotates clockwise on the turntable."
|
| 6 |
+
},
|
| 7 |
+
"camera_info": {
|
| 8 |
+
"color": "Deep brown, amber, green pistachio, gold highlights",
|
| 9 |
+
"frame_size": "Extreme close up",
|
| 10 |
+
"shot_type_angle": "Slight high angle",
|
| 11 |
+
"lens_size": "Macro lens",
|
| 12 |
+
"composition": "Dessert centered with glaze entering from top",
|
| 13 |
+
"lighting": "Large softbox reflections",
|
| 14 |
+
"lighting_type": "Controlled studio kitchen lighting"
|
| 15 |
+
},
|
| 16 |
+
"world_knowledge": [],
|
| 17 |
+
"prominent_elements": [
|
| 18 |
+
{
|
| 19 |
+
"name": "chocolate dome dessert",
|
| 20 |
+
"description": "A small mirror-glazed chocolate dome on a black plate, surrounded by pistachio dust and gold flakes.",
|
| 21 |
+
"actions": [
|
| 22 |
+
{
|
| 23 |
+
"timestamp": "[0.0s - 2.0s]",
|
| 24 |
+
"action": "rotates slowly clockwise on a metal turntable"
|
| 25 |
+
}
|
| 26 |
+
],
|
| 27 |
+
"location": "center of the frame",
|
| 28 |
+
"relative_size": "dominant",
|
| 29 |
+
"shape_and_color": "hemispherical dark chocolate dome on round black plate",
|
| 30 |
+
"texture": "mirror-gloss surface with smooth curved reflections",
|
| 31 |
+
"appearance_details": "tiny gold flakes, green pistachio crumbs, clean pastry plating",
|
| 32 |
+
"relationship": "receives the caramel glaze",
|
| 33 |
+
"orientation": "upright on the turntable",
|
| 34 |
+
"pose": "",
|
| 35 |
+
"expression": "",
|
| 36 |
+
"clothing": "",
|
| 37 |
+
"is_cluster": false,
|
| 38 |
+
"number_of_objects": ""
|
| 39 |
+
},
|
| 40 |
+
{
|
| 41 |
+
"name": "caramel glaze stream",
|
| 42 |
+
"description": "A viscous amber caramel stream pouring from above and spreading over the dessert.",
|
| 43 |
+
"actions": [
|
| 44 |
+
{
|
| 45 |
+
"timestamp": "[0.0s - 0.8s]",
|
| 46 |
+
"action": "falls as a narrow glossy stream onto the crown of the dome"
|
| 47 |
+
},
|
| 48 |
+
{
|
| 49 |
+
"timestamp": "[0.8s - 1.6s]",
|
| 50 |
+
"action": "forms slow ribbons that slide down the curved chocolate surface"
|
| 51 |
+
},
|
| 52 |
+
{
|
| 53 |
+
"timestamp": "[1.6s - 2.0s]",
|
| 54 |
+
"action": "collects at the base in a thin shiny ring"
|
| 55 |
+
}
|
| 56 |
+
],
|
| 57 |
+
"location": "entering from top center and flowing over the dessert",
|
| 58 |
+
"relative_size": "medium",
|
| 59 |
+
"shape_and_color": "thin amber liquid ribbons",
|
| 60 |
+
"texture": "viscous, glossy, translucent caramel",
|
| 61 |
+
"appearance_details": "bright highlights and tiny bubbles",
|
| 62 |
+
"relationship": "coats the chocolate dome",
|
| 63 |
+
"orientation": "vertical stream becoming downward trails",
|
| 64 |
+
"pose": "",
|
| 65 |
+
"expression": "",
|
| 66 |
+
"clothing": "",
|
| 67 |
+
"is_cluster": false,
|
| 68 |
+
"number_of_objects": ""
|
| 69 |
+
}
|
| 70 |
+
]
|
| 71 |
+
},
|
| 72 |
+
"duration": 2.0
|
| 73 |
+
}
|
benchmark/prompts/03_rain_market_lanterns.json
ADDED
|
@@ -0,0 +1,99 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"caption": {
|
| 3 |
+
"comprehensive_description": {
|
| 4 |
+
"scene_content_description": "A lively rainy night market in a narrow old-city street, filmed like naturalistic cinema. Red paper lanterns hang from wooden stalls, steam rises from bowls of soup, and wet cobblestones mirror orange and teal reflections. A vendor in a dark apron slides a steaming bowl across a counter to a traveler in a blue raincoat. Umbrellas pass in the foreground, partially occluding the scene for a moment. The mood is warm, humid, detailed, and grounded, with no readable signs or subtitles.",
|
| 5 |
+
"camera_movement_description": "The camera performs a slow lateral dolly from left to right at counter height, with foreground umbrellas briefly crossing the lens."
|
| 6 |
+
},
|
| 7 |
+
"camera_info": {
|
| 8 |
+
"color": "Warm lantern orange, teal rain reflections, red accents",
|
| 9 |
+
"frame_size": "Medium wide",
|
| 10 |
+
"shot_type_angle": "Eye level",
|
| 11 |
+
"lens_size": "Medium lens",
|
| 12 |
+
"composition": "Vendor stall centered, traveler on right, passing umbrellas in foreground",
|
| 13 |
+
"lighting": "Lantern light mixed with rain reflections",
|
| 14 |
+
"lighting_type": "Night practical lighting"
|
| 15 |
+
},
|
| 16 |
+
"world_knowledge": [],
|
| 17 |
+
"prominent_elements": [
|
| 18 |
+
{
|
| 19 |
+
"name": "food vendor",
|
| 20 |
+
"description": "A middle-aged street vendor wearing a dark apron and working behind a steaming wooden food stall.",
|
| 21 |
+
"actions": [
|
| 22 |
+
{
|
| 23 |
+
"timestamp": "[0.0s - 0.8s]",
|
| 24 |
+
"action": "lifts a white ceramic bowl from a steaming pot"
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"timestamp": "[0.8s - 1.6s]",
|
| 28 |
+
"action": "slides the bowl carefully across the counter"
|
| 29 |
+
},
|
| 30 |
+
{
|
| 31 |
+
"timestamp": "[1.6s - 2.0s]",
|
| 32 |
+
"action": "wipes condensation from the counter with a cloth"
|
| 33 |
+
}
|
| 34 |
+
],
|
| 35 |
+
"location": "center-left behind the stall counter",
|
| 36 |
+
"relative_size": "large",
|
| 37 |
+
"shape_and_color": "human figure in dark apron under warm lantern light",
|
| 38 |
+
"texture": "wet fabric, skin highlights, wood grain around the stall",
|
| 39 |
+
"appearance_details": "rolled sleeves, focused expression, steam around hands",
|
| 40 |
+
"relationship": "serves food to the traveler",
|
| 41 |
+
"orientation": "facing right toward the customer",
|
| 42 |
+
"pose": "leaning slightly forward over the counter",
|
| 43 |
+
"expression": "focused and calm",
|
| 44 |
+
"clothing": "dark apron over a muted shirt",
|
| 45 |
+
"gender": "",
|
| 46 |
+
"skin_tone_and_texture": "natural skin with rain-lit highlights"
|
| 47 |
+
},
|
| 48 |
+
{
|
| 49 |
+
"name": "traveler in blue raincoat",
|
| 50 |
+
"description": "A traveler wearing a blue hooded raincoat, waiting at the stall with wet sleeves and a folded umbrella.",
|
| 51 |
+
"actions": [
|
| 52 |
+
{
|
| 53 |
+
"timestamp": "[0.0s - 1.2s]",
|
| 54 |
+
"action": "waits with hands near the counter, watching the bowl"
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"timestamp": "[1.2s - 2.0s]",
|
| 58 |
+
"action": "reaches forward to receive the steaming bowl"
|
| 59 |
+
}
|
| 60 |
+
],
|
| 61 |
+
"location": "right side of the frame",
|
| 62 |
+
"relative_size": "medium",
|
| 63 |
+
"shape_and_color": "blue hooded raincoat silhouette",
|
| 64 |
+
"texture": "water beads on waterproof fabric",
|
| 65 |
+
"appearance_details": "hood up, folded umbrella tucked under one arm",
|
| 66 |
+
"relationship": "customer receiving food from the vendor",
|
| 67 |
+
"orientation": "facing left toward the stall",
|
| 68 |
+
"pose": "standing close to the counter",
|
| 69 |
+
"expression": "expectant and tired",
|
| 70 |
+
"clothing": "blue raincoat",
|
| 71 |
+
"gender": "",
|
| 72 |
+
"skin_tone_and_texture": ""
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"name": "rain and lantern reflections",
|
| 76 |
+
"description": "Falling rain, steam, and glowing reflections on wet cobblestones.",
|
| 77 |
+
"actions": [
|
| 78 |
+
{
|
| 79 |
+
"timestamp": "[0.0s - 2.0s]",
|
| 80 |
+
"action": "rain falls continuously while reflections ripple on the street"
|
| 81 |
+
}
|
| 82 |
+
],
|
| 83 |
+
"location": "foreground and background throughout the frame",
|
| 84 |
+
"relative_size": "large",
|
| 85 |
+
"shape_and_color": "thin rain streaks, orange and teal reflected pools",
|
| 86 |
+
"texture": "wet stone, vapor, glossy water",
|
| 87 |
+
"appearance_details": "soft steam plumes, umbrellas passing close to lens",
|
| 88 |
+
"relationship": "sets the atmosphere around the market",
|
| 89 |
+
"orientation": "vertical rainfall and horizontal street reflections",
|
| 90 |
+
"pose": "",
|
| 91 |
+
"expression": "",
|
| 92 |
+
"clothing": "",
|
| 93 |
+
"is_cluster": true,
|
| 94 |
+
"number_of_objects": "many raindrops and reflections"
|
| 95 |
+
}
|
| 96 |
+
]
|
| 97 |
+
},
|
| 98 |
+
"duration": 2.0
|
| 99 |
+
}
|
benchmark/prompts/04_underwater_manta.json
ADDED
|
@@ -0,0 +1,95 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"caption": {
|
| 3 |
+
"comprehensive_description": {
|
| 4 |
+
"scene_content_description": "A tranquil underwater wildlife scene in clear tropical water. A large manta ray glides above a colorful coral garden while a scuba diver hovers several meters behind, holding a small camera rig. Sunbeams ripple through the surface, schools of tiny silver fish split and rejoin around the manta, and suspended particles drift slowly. The image should feel like realistic underwater documentary footage with natural motion, soft blue-green color, and no fantasy elements.",
|
| 5 |
+
"camera_movement_description": "The camera tracks smoothly beside the manta ray from a slightly lower angle, drifting forward with the current."
|
| 6 |
+
},
|
| 7 |
+
"camera_info": {
|
| 8 |
+
"color": "Blue-green water with coral reds and yellows",
|
| 9 |
+
"frame_size": "Wide",
|
| 10 |
+
"shot_type_angle": "Slight low angle",
|
| 11 |
+
"lens_size": "Wide underwater lens",
|
| 12 |
+
"composition": "Manta ray crossing the upper center, coral below, diver in background",
|
| 13 |
+
"lighting": "Dappled sunlight",
|
| 14 |
+
"lighting_type": "Natural underwater daylight"
|
| 15 |
+
},
|
| 16 |
+
"world_knowledge": [],
|
| 17 |
+
"prominent_elements": [
|
| 18 |
+
{
|
| 19 |
+
"name": "manta ray",
|
| 20 |
+
"description": "A large manta ray with broad triangular fins, dark top surface, and pale underside.",
|
| 21 |
+
"actions": [
|
| 22 |
+
{
|
| 23 |
+
"timestamp": "[0.0s - 0.8s]",
|
| 24 |
+
"action": "glides from left to right with slow wing-like fin movement"
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"timestamp": "[0.8s - 1.5s]",
|
| 28 |
+
"action": "tilts slightly upward as sunbeams cross its back"
|
| 29 |
+
},
|
| 30 |
+
{
|
| 31 |
+
"timestamp": "[1.5s - 2.0s]",
|
| 32 |
+
"action": "continues forward while small fish scatter around it"
|
| 33 |
+
}
|
| 34 |
+
],
|
| 35 |
+
"location": "upper center moving toward right",
|
| 36 |
+
"relative_size": "dominant",
|
| 37 |
+
"shape_and_color": "wide diamond-like silhouette, dark grey top, pale underside",
|
| 38 |
+
"texture": "smooth skin with subtle mottling",
|
| 39 |
+
"appearance_details": "cephalic fins near mouth, long tail trailing behind",
|
| 40 |
+
"relationship": "main wildlife subject of the video",
|
| 41 |
+
"orientation": "moving left to right and slightly upward",
|
| 42 |
+
"pose": "",
|
| 43 |
+
"expression": "",
|
| 44 |
+
"clothing": "",
|
| 45 |
+
"is_cluster": false,
|
| 46 |
+
"number_of_objects": ""
|
| 47 |
+
},
|
| 48 |
+
{
|
| 49 |
+
"name": "scuba diver",
|
| 50 |
+
"description": "A diver in black wetsuit and fins, carrying a small underwater camera rig.",
|
| 51 |
+
"actions": [
|
| 52 |
+
{
|
| 53 |
+
"timestamp": "[0.0s - 2.0s]",
|
| 54 |
+
"action": "hovers calmly in the background, exhaling small bubble streams"
|
| 55 |
+
}
|
| 56 |
+
],
|
| 57 |
+
"location": "mid-background behind the manta ray",
|
| 58 |
+
"relative_size": "small",
|
| 59 |
+
"shape_and_color": "human figure in black wetsuit with silver air tank",
|
| 60 |
+
"texture": "neoprene suit, metal tank, glass mask",
|
| 61 |
+
"appearance_details": "fins, mask, regulator, compact camera lights",
|
| 62 |
+
"relationship": "observes the manta ray without touching it",
|
| 63 |
+
"orientation": "facing the manta ray",
|
| 64 |
+
"pose": "horizontal hover",
|
| 65 |
+
"expression": "",
|
| 66 |
+
"clothing": "black scuba gear",
|
| 67 |
+
"gender": "",
|
| 68 |
+
"skin_tone_and_texture": ""
|
| 69 |
+
},
|
| 70 |
+
{
|
| 71 |
+
"name": "coral reef and fish",
|
| 72 |
+
"description": "A colorful coral garden with small silver fish schooling above it.",
|
| 73 |
+
"actions": [
|
| 74 |
+
{
|
| 75 |
+
"timestamp": "[0.0s - 2.0s]",
|
| 76 |
+
"action": "fish swirl and split around the manta while coral remains fixed"
|
| 77 |
+
}
|
| 78 |
+
],
|
| 79 |
+
"location": "lower half of the frame",
|
| 80 |
+
"relative_size": "large",
|
| 81 |
+
"shape_and_color": "branching corals in red, yellow, and muted purple; silver fish clusters",
|
| 82 |
+
"texture": "rough coral, shimmering fish scales",
|
| 83 |
+
"appearance_details": "sand patches, sea fans, drifting particles",
|
| 84 |
+
"relationship": "environment below the manta ray",
|
| 85 |
+
"orientation": "reef spreads horizontally across the bottom",
|
| 86 |
+
"pose": "",
|
| 87 |
+
"expression": "",
|
| 88 |
+
"clothing": "",
|
| 89 |
+
"is_cluster": true,
|
| 90 |
+
"number_of_objects": "many fish and coral structures"
|
| 91 |
+
}
|
| 92 |
+
]
|
| 93 |
+
},
|
| 94 |
+
"duration": 2.0
|
| 95 |
+
}
|
benchmark/prompts/05_factory_battery_cells.json
ADDED
|
@@ -0,0 +1,111 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"caption": {
|
| 3 |
+
"comprehensive_description": {
|
| 4 |
+
"scene_content_description": "A clean high-tech factory line assembling transparent solid-state battery cells. Two white robotic arms move with precise synchronized motion over a brushed steel conveyor. One arm lowers a translucent rectangular cell into a copper test fixture while the other arm scans it with a blue light bar. Tiny reflections slide across glass safety panels, and status lights pulse softly. The scene should look like realistic industrial automation footage, not CGI, with crisp metal surfaces and controlled motion.",
|
| 5 |
+
"camera_movement_description": "The camera is locked off in a medium-wide three-quarter view with a subtle mechanical vibration, emphasizing repeatable robotic motion."
|
| 6 |
+
},
|
| 7 |
+
"camera_info": {
|
| 8 |
+
"color": "White, brushed steel, copper, blue inspection light",
|
| 9 |
+
"frame_size": "Medium wide",
|
| 10 |
+
"shot_type_angle": "Three-quarter eye-level industrial view",
|
| 11 |
+
"lens_size": "Medium lens",
|
| 12 |
+
"composition": "Two robotic arms framing the battery cell at center",
|
| 13 |
+
"lighting": "Even overhead factory lighting with blue scanner highlight",
|
| 14 |
+
"lighting_type": "Artificial industrial lighting"
|
| 15 |
+
},
|
| 16 |
+
"world_knowledge": [],
|
| 17 |
+
"prominent_elements": [
|
| 18 |
+
{
|
| 19 |
+
"name": "left robotic arm",
|
| 20 |
+
"description": "A white six-axis robotic arm with a vacuum gripper holding a transparent rectangular battery cell.",
|
| 21 |
+
"actions": [
|
| 22 |
+
{
|
| 23 |
+
"timestamp": "[0.0s - 0.7s]",
|
| 24 |
+
"action": "moves downward from upper left with the transparent cell held steady"
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"timestamp": "[0.7s - 1.4s]",
|
| 28 |
+
"action": "places the cell into a copper test fixture"
|
| 29 |
+
},
|
| 30 |
+
{
|
| 31 |
+
"timestamp": "[1.4s - 2.0s]",
|
| 32 |
+
"action": "releases the cell and retracts slightly upward"
|
| 33 |
+
}
|
| 34 |
+
],
|
| 35 |
+
"location": "left side moving toward center",
|
| 36 |
+
"relative_size": "large",
|
| 37 |
+
"shape_and_color": "white articulated segments with black joints",
|
| 38 |
+
"texture": "smooth painted metal and rubber vacuum cups",
|
| 39 |
+
"appearance_details": "visible cable routing, small green status LED",
|
| 40 |
+
"relationship": "positions the battery cell for testing",
|
| 41 |
+
"orientation": "angled downward toward the center fixture",
|
| 42 |
+
"pose": "",
|
| 43 |
+
"expression": "",
|
| 44 |
+
"clothing": "",
|
| 45 |
+
"is_cluster": false,
|
| 46 |
+
"number_of_objects": ""
|
| 47 |
+
},
|
| 48 |
+
{
|
| 49 |
+
"name": "right robotic scanner arm",
|
| 50 |
+
"description": "A second white robotic arm carrying a rectangular blue inspection light bar.",
|
| 51 |
+
"actions": [
|
| 52 |
+
{
|
| 53 |
+
"timestamp": "[0.0s - 0.9s]",
|
| 54 |
+
"action": "waits above the copper fixture with blue scanner light dim"
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"timestamp": "[0.9s - 1.6s]",
|
| 58 |
+
"action": "sweeps the blue light bar across the transparent cell"
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"timestamp": "[1.6s - 2.0s]",
|
| 62 |
+
"action": "pauses as the scanner light pulses once"
|
| 63 |
+
}
|
| 64 |
+
],
|
| 65 |
+
"location": "right side above the conveyor",
|
| 66 |
+
"relative_size": "large",
|
| 67 |
+
"shape_and_color": "white arm with glowing blue rectangular scanner",
|
| 68 |
+
"texture": "smooth metal casing and glass scanner cover",
|
| 69 |
+
"appearance_details": "blue light strip, black joints, compact sensor module",
|
| 70 |
+
"relationship": "inspects the cell after placement",
|
| 71 |
+
"orientation": "angled left toward the battery cell",
|
| 72 |
+
"pose": "",
|
| 73 |
+
"expression": "",
|
| 74 |
+
"clothing": "",
|
| 75 |
+
"is_cluster": false,
|
| 76 |
+
"number_of_objects": ""
|
| 77 |
+
},
|
| 78 |
+
{
|
| 79 |
+
"name": "transparent battery cell and copper fixture",
|
| 80 |
+
"description": "A clear rectangular solid-state cell seated into a copper test fixture on a steel conveyor.",
|
| 81 |
+
"actions": [
|
| 82 |
+
{
|
| 83 |
+
"timestamp": "[0.0s - 0.7s]",
|
| 84 |
+
"action": "moves with the left arm above the fixture"
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"timestamp": "[0.7s - 1.4s]",
|
| 88 |
+
"action": "is lowered precisely into the fixture"
|
| 89 |
+
},
|
| 90 |
+
{
|
| 91 |
+
"timestamp": "[1.4s - 2.0s]",
|
| 92 |
+
"action": "remains stationary while blue inspection light passes over it"
|
| 93 |
+
}
|
| 94 |
+
],
|
| 95 |
+
"location": "center of the frame on the conveyor",
|
| 96 |
+
"relative_size": "medium",
|
| 97 |
+
"shape_and_color": "transparent rectangle with faint internal layers, copper fixture",
|
| 98 |
+
"texture": "clear glass-like cell, brushed copper, polished steel",
|
| 99 |
+
"appearance_details": "thin internal laminate layers, small alignment pins, clean conveyor surface",
|
| 100 |
+
"relationship": "object being assembled and inspected",
|
| 101 |
+
"orientation": "horizontal in the fixture",
|
| 102 |
+
"pose": "",
|
| 103 |
+
"expression": "",
|
| 104 |
+
"clothing": "",
|
| 105 |
+
"is_cluster": false,
|
| 106 |
+
"number_of_objects": ""
|
| 107 |
+
}
|
| 108 |
+
]
|
| 109 |
+
},
|
| 110 |
+
"duration": 2.0
|
| 111 |
+
}
|
benchmark/refiner.original.metrics.csv
ADDED
|
@@ -0,0 +1,2 @@
|
|
|
|
|
|
|
|
|
|
| 1 |
+
base_video,base_video_sha256,batch_cfg,ffprobe,fps,generation_seconds,guidance_scale,height,model_key,negative_prompt,null_cond_clone_zero,num_frames,num_inference_steps,nvidia_smi_after,output_file,output_sha256,output_size_bytes,peak_disk_used_bytes,peak_gpu_memory_used_mb,peak_gpu_power_watts,peak_gpu_utilization_percent,peak_process_rss_bytes,peak_system_used_bytes,prompt_id,raw_frames_file,raw_frames_sha256,sample_count,sample_interval_seconds,seed,shared_initial_latent,shared_initial_latent_sha256,shift,sigma_tail_steps,t_thresh,torch_allocated_after_mb,torch_peak_allocated_mb,torch_peak_reserved_mb,torch_reserved_after_mb,variant,width
|
| 2 |
+
results/benchmark/moe/original/videos/01_arctic_research_drone.mp4,e84547fdbd70c1f118435c0b0dee74ccb719a64f1c51e2d7b1260f2185f3dad9,False,"{'programs': [], 'streams': [{'codec_name': 'h264', 'duration': '3.041667', 'height': 1088, 'nb_frames': '73', 'pix_fmt': 'yuv420p', 'r_frame_rate': '24/1', 'width': 1920}]}",24,478.6046418759506,3.0,1088,moe_refiner,"{""universal_negative"": {""visual_quality"": [""low quality"", ""worst quality"", ""blurry"", ""pixelated"", ""jpeg artifacts"", ""low resolution"", ""unstable color"", ""color flicker"", ""underexposed"", ""overexposed"", ""invisible subject"", ""subject hidden in darkness""], ""artistic_style"": [""painting"", ""illustration"", ""drawing"", ""cartoon"", ""3d render"", ""cgi"", ""sketch"", ""digital art""], ""composition_and_content"": [""text"", ""watermark"", ""signature"", ""logo"", ""subtitles"", ""pillarboxed"", ""side bars"", ""portrait image in landscape frame""], ""temporal_and_motion_stability"": [""flickering"", ""jittery"", ""motion blur"", ""temporal inconsistency"", ""warping"", ""morphing"", ""incoherent motion"", ""unnatural movement"", ""static object with sudden jump"", ""frame-to-frame inconsistency""], ""material_and_structure"": [""plastic-like glass"", ""unrealistic texture"", ""deformed bottle"", ""liquid freezing improperly"", ""distorted reflections""]}}",True,73,8,"{'memory.total': '183359', 'memory.used': '179178', 'name': 'NVIDIA B200', 'power.draw': '252.18', 'temperature.gpu': '35', 'timestamp': '2026/07/10 09:42:11.197', 'utilization.gpu': '0'}",results/benchmark/moe_refiner/original/videos/01_arctic_research_drone.mp4,b3bdfa0eaa7f4a3b9cbc3627f15f144b2416778f5160e3242bb2fc1ff284a7de,1953653,232971304960.0,179178.0,965.67,100.0,4571336704.0,115570741248.0,01_arctic_research_drone,results/benchmark/moe_refiner/original/raw_frames/01_arctic_research_drone.npy,36469b4d213a6e5eeaaa3ad49005fbb6a9c7417c598f1d251b7e0fedc1ed514a,1096,0.25,4201,results/benchmark/moe/refiner_shared/01_arctic_research_drone.safetensors,502b10f841d96aa101e69421b20083aeb60427c054b42fb0db29c5a6e70824cf,3.0,2,0.85,103627.8291015625,150526.1474609375,178056.0,178056.0,original,1920
|
benchmark/refiner.original.metrics.jsonl
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"base_video": "results/benchmark/moe/original/videos/01_arctic_research_drone.mp4", "base_video_sha256": "e84547fdbd70c1f118435c0b0dee74ccb719a64f1c51e2d7b1260f2185f3dad9", "batch_cfg": false, "ffprobe": {"programs": [], "streams": [{"codec_name": "h264", "duration": "3.041667", "height": 1088, "nb_frames": "73", "pix_fmt": "yuv420p", "r_frame_rate": "24/1", "width": 1920}]}, "fps": 24, "generation_seconds": 478.6046418759506, "guidance_scale": 3.0, "height": 1088, "model_key": "moe_refiner", "negative_prompt": "{\"universal_negative\": {\"visual_quality\": [\"low quality\", \"worst quality\", \"blurry\", \"pixelated\", \"jpeg artifacts\", \"low resolution\", \"unstable color\", \"color flicker\", \"underexposed\", \"overexposed\", \"invisible subject\", \"subject hidden in darkness\"], \"artistic_style\": [\"painting\", \"illustration\", \"drawing\", \"cartoon\", \"3d render\", \"cgi\", \"sketch\", \"digital art\"], \"composition_and_content\": [\"text\", \"watermark\", \"signature\", \"logo\", \"subtitles\", \"pillarboxed\", \"side bars\", \"portrait image in landscape frame\"], \"temporal_and_motion_stability\": [\"flickering\", \"jittery\", \"motion blur\", \"temporal inconsistency\", \"warping\", \"morphing\", \"incoherent motion\", \"unnatural movement\", \"static object with sudden jump\", \"frame-to-frame inconsistency\"], \"material_and_structure\": [\"plastic-like glass\", \"unrealistic texture\", \"deformed bottle\", \"liquid freezing improperly\", \"distorted reflections\"]}}", "null_cond_clone_zero": true, "num_frames": 73, "num_inference_steps": 8, "nvidia_smi_after": {"memory.total": "183359", "memory.used": "179178", "name": "NVIDIA B200", "power.draw": "252.18", "temperature.gpu": "35", "timestamp": "2026/07/10 09:42:11.197", "utilization.gpu": "0"}, "output_file": "results/benchmark/moe_refiner/original/videos/01_arctic_research_drone.mp4", "output_sha256": "b3bdfa0eaa7f4a3b9cbc3627f15f144b2416778f5160e3242bb2fc1ff284a7de", "output_size_bytes": 1953653, "peak_disk_used_bytes": 232971304960.0, "peak_gpu_memory_used_mb": 179178.0, "peak_gpu_power_watts": 965.67, "peak_gpu_utilization_percent": 100.0, "peak_process_rss_bytes": 4571336704.0, "peak_system_used_bytes": 115570741248.0, "prompt_id": "01_arctic_research_drone", "raw_frames_file": "results/benchmark/moe_refiner/original/raw_frames/01_arctic_research_drone.npy", "raw_frames_sha256": "36469b4d213a6e5eeaaa3ad49005fbb6a9c7417c598f1d251b7e0fedc1ed514a", "sample_count": 1096, "sample_interval_seconds": 0.25, "seed": 4201, "shared_initial_latent": "results/benchmark/moe/refiner_shared/01_arctic_research_drone.safetensors", "shared_initial_latent_sha256": "502b10f841d96aa101e69421b20083aeb60427c054b42fb0db29c5a6e70824cf", "shift": 3.0, "sigma_tail_steps": 2, "t_thresh": 0.85, "torch_allocated_after_mb": 103627.8291015625, "torch_peak_allocated_mb": 150526.1474609375, "torch_peak_reserved_mb": 178056.0, "torch_reserved_after_mb": 178056.0, "variant": "original", "width": 1920}
|
benchmark/refiner.paired.metrics.csv
ADDED
|
@@ -0,0 +1,2 @@
|
|
|
|
|
|
|
|
|
|
| 1 |
+
prompt_id,seed,height,width,num_frames,fps,num_inference_steps,guidance_scale,shift,original_generation_seconds,original_seconds_per_frame,original_torch_peak_allocated_mb,original_torch_peak_reserved_mb,original_peak_gpu_memory_used_mb,original_peak_gpu_utilization_percent,original_peak_process_rss_bytes,original_peak_system_used_bytes,original_output_size_bytes,original_output_sha256,sdnq_generation_seconds,sdnq_seconds_per_frame,sdnq_torch_peak_allocated_mb,sdnq_torch_peak_reserved_mb,sdnq_peak_gpu_memory_used_mb,sdnq_peak_gpu_utilization_percent,sdnq_peak_process_rss_bytes,sdnq_peak_system_used_bytes,sdnq_output_size_bytes,sdnq_output_sha256
|
| 2 |
+
01_arctic_research_drone,4201,1088,1920,73,24,8,3.0,3.0,478.6046418759506,,150526.1474609375,178056.0,179178.0,100.0,4571336704.0,115570741248.0,1953653,b3bdfa0eaa7f4a3b9cbc3627f15f144b2416778f5160e3242bb2fc1ff284a7de,460.47301308601163,,74121.7998046875,97110.0,98232.0,100.0,4135526400.0,113527255040.0,1910664,fb95a8d0cda0c3038e107a53eb3bc5b1df840a8568318c6f90bb4d0566a7674a
|
benchmark/refiner.paired.metrics.jsonl
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"fps": 24, "guidance_scale": 3.0, "height": 1088, "num_frames": 73, "num_inference_steps": 8, "original_generation_seconds": 478.6046418759506, "original_output_sha256": "b3bdfa0eaa7f4a3b9cbc3627f15f144b2416778f5160e3242bb2fc1ff284a7de", "original_output_size_bytes": 1953653, "original_peak_gpu_memory_used_mb": 179178.0, "original_peak_gpu_utilization_percent": 100.0, "original_peak_process_rss_bytes": 4571336704.0, "original_peak_system_used_bytes": 115570741248.0, "original_seconds_per_frame": null, "original_torch_peak_allocated_mb": 150526.1474609375, "original_torch_peak_reserved_mb": 178056.0, "prompt_id": "01_arctic_research_drone", "sdnq_generation_seconds": 460.47301308601163, "sdnq_output_sha256": "fb95a8d0cda0c3038e107a53eb3bc5b1df840a8568318c6f90bb4d0566a7674a", "sdnq_output_size_bytes": 1910664, "sdnq_peak_gpu_memory_used_mb": 98232.0, "sdnq_peak_gpu_utilization_percent": 100.0, "sdnq_peak_process_rss_bytes": 4135526400.0, "sdnq_peak_system_used_bytes": 113527255040.0, "sdnq_seconds_per_frame": null, "sdnq_torch_peak_allocated_mb": 74121.7998046875, "sdnq_torch_peak_reserved_mb": 97110.0, "seed": 4201, "shift": 3.0, "width": 1920}
|
benchmark/refiner.sdnq.metrics.csv
ADDED
|
@@ -0,0 +1,2 @@
|
|
|
|
|
|
|
|
|
|
| 1 |
+
base_video,base_video_sha256,batch_cfg,ffprobe,fps,generation_seconds,guidance_scale,height,model_key,negative_prompt,null_cond_clone_zero,num_frames,num_inference_steps,nvidia_smi_after,output_file,output_sha256,output_size_bytes,peak_disk_used_bytes,peak_gpu_memory_used_mb,peak_gpu_power_watts,peak_gpu_utilization_percent,peak_process_rss_bytes,peak_system_used_bytes,prompt_id,raw_frames_file,raw_frames_sha256,sample_count,sample_interval_seconds,seed,shared_initial_latent,shared_initial_latent_sha256,shift,sigma_tail_steps,t_thresh,torch_allocated_after_mb,torch_peak_allocated_mb,torch_peak_reserved_mb,torch_reserved_after_mb,variant,width
|
| 2 |
+
results/benchmark/moe/original/videos/01_arctic_research_drone.mp4,e84547fdbd70c1f118435c0b0dee74ccb719a64f1c51e2d7b1260f2185f3dad9,False,"{'programs': [], 'streams': [{'codec_name': 'h264', 'duration': '3.041667', 'height': 1088, 'nb_frames': '73', 'pix_fmt': 'yuv420p', 'r_frame_rate': '24/1', 'width': 1920}]}",24,460.47301308601163,3.0,1088,moe_refiner,"{""universal_negative"": {""visual_quality"": [""low quality"", ""worst quality"", ""blurry"", ""pixelated"", ""jpeg artifacts"", ""low resolution"", ""unstable color"", ""color flicker"", ""underexposed"", ""overexposed"", ""invisible subject"", ""subject hidden in darkness""], ""artistic_style"": [""painting"", ""illustration"", ""drawing"", ""cartoon"", ""3d render"", ""cgi"", ""sketch"", ""digital art""], ""composition_and_content"": [""text"", ""watermark"", ""signature"", ""logo"", ""subtitles"", ""pillarboxed"", ""side bars"", ""portrait image in landscape frame""], ""temporal_and_motion_stability"": [""flickering"", ""jittery"", ""motion blur"", ""temporal inconsistency"", ""warping"", ""morphing"", ""incoherent motion"", ""unnatural movement"", ""static object with sudden jump"", ""frame-to-frame inconsistency""], ""material_and_structure"": [""plastic-like glass"", ""unrealistic texture"", ""deformed bottle"", ""liquid freezing improperly"", ""distorted reflections""]}}",True,73,8,"{'memory.total': '183359', 'memory.used': '98232', 'name': 'NVIDIA B200', 'power.draw': '255.56', 'temperature.gpu': '38', 'timestamp': '2026/07/10 09:53:13.523', 'utilization.gpu': '0'}",results/benchmark/moe_refiner/sdnq/videos/01_arctic_research_drone.mp4,fb95a8d0cda0c3038e107a53eb3bc5b1df840a8568318c6f90bb4d0566a7674a,1910664,231090356224.0,98232.0,971.84,100.0,4135526400.0,113527255040.0,01_arctic_research_drone,results/benchmark/moe_refiner/sdnq/raw_frames/01_arctic_research_drone.npy,5e4e476621672882660e15a223f1e0ab5812be1cdc37639263f5a476142e4d89,1477,0.25,4201,results/benchmark/moe/refiner_shared/01_arctic_research_drone.safetensors,502b10f841d96aa101e69421b20083aeb60427c054b42fb0db29c5a6e70824cf,3.0,2,0.85,27224.1650390625,74121.7998046875,97110.0,97110.0,sdnq,1920
|
benchmark/refiner.sdnq.metrics.jsonl
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"base_video": "results/benchmark/moe/original/videos/01_arctic_research_drone.mp4", "base_video_sha256": "e84547fdbd70c1f118435c0b0dee74ccb719a64f1c51e2d7b1260f2185f3dad9", "batch_cfg": false, "ffprobe": {"programs": [], "streams": [{"codec_name": "h264", "duration": "3.041667", "height": 1088, "nb_frames": "73", "pix_fmt": "yuv420p", "r_frame_rate": "24/1", "width": 1920}]}, "fps": 24, "generation_seconds": 460.47301308601163, "guidance_scale": 3.0, "height": 1088, "model_key": "moe_refiner", "negative_prompt": "{\"universal_negative\": {\"visual_quality\": [\"low quality\", \"worst quality\", \"blurry\", \"pixelated\", \"jpeg artifacts\", \"low resolution\", \"unstable color\", \"color flicker\", \"underexposed\", \"overexposed\", \"invisible subject\", \"subject hidden in darkness\"], \"artistic_style\": [\"painting\", \"illustration\", \"drawing\", \"cartoon\", \"3d render\", \"cgi\", \"sketch\", \"digital art\"], \"composition_and_content\": [\"text\", \"watermark\", \"signature\", \"logo\", \"subtitles\", \"pillarboxed\", \"side bars\", \"portrait image in landscape frame\"], \"temporal_and_motion_stability\": [\"flickering\", \"jittery\", \"motion blur\", \"temporal inconsistency\", \"warping\", \"morphing\", \"incoherent motion\", \"unnatural movement\", \"static object with sudden jump\", \"frame-to-frame inconsistency\"], \"material_and_structure\": [\"plastic-like glass\", \"unrealistic texture\", \"deformed bottle\", \"liquid freezing improperly\", \"distorted reflections\"]}}", "null_cond_clone_zero": true, "num_frames": 73, "num_inference_steps": 8, "nvidia_smi_after": {"memory.total": "183359", "memory.used": "98232", "name": "NVIDIA B200", "power.draw": "255.56", "temperature.gpu": "38", "timestamp": "2026/07/10 09:53:13.523", "utilization.gpu": "0"}, "output_file": "results/benchmark/moe_refiner/sdnq/videos/01_arctic_research_drone.mp4", "output_sha256": "fb95a8d0cda0c3038e107a53eb3bc5b1df840a8568318c6f90bb4d0566a7674a", "output_size_bytes": 1910664, "peak_disk_used_bytes": 231090356224.0, "peak_gpu_memory_used_mb": 98232.0, "peak_gpu_power_watts": 971.84, "peak_gpu_utilization_percent": 100.0, "peak_process_rss_bytes": 4135526400.0, "peak_system_used_bytes": 113527255040.0, "prompt_id": "01_arctic_research_drone", "raw_frames_file": "results/benchmark/moe_refiner/sdnq/raw_frames/01_arctic_research_drone.npy", "raw_frames_sha256": "5e4e476621672882660e15a223f1e0ab5812be1cdc37639263f5a476142e4d89", "sample_count": 1477, "sample_interval_seconds": 0.25, "seed": 4201, "shared_initial_latent": "results/benchmark/moe/refiner_shared/01_arctic_research_drone.safetensors", "shared_initial_latent_sha256": "502b10f841d96aa101e69421b20083aeb60427c054b42fb0db29c5a6e70824cf", "shift": 3.0, "sigma_tail_steps": 2, "t_thresh": 0.85, "torch_allocated_after_mb": 27224.1650390625, "torch_peak_allocated_mb": 74121.7998046875, "torch_peak_reserved_mb": 97110.0, "torch_reserved_after_mb": 97110.0, "variant": "sdnq", "width": 1920}
|
benchmark/sdnq.metrics.csv
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
model_key,variant,prompt_id,seed,generation_seconds,seconds_per_frame,torch_peak_allocated_mb,torch_peak_reserved_mb,peak_gpu_memory_used_mb,peak_gpu_utilization_percent,peak_process_rss_bytes,peak_system_used_bytes,output_size_bytes,output_sha256
|
| 2 |
+
moe,sdnq,01_arctic_research_drone,4201,159.77318301890045,2.188673739984938,36067.29833984375,40756.0,41878.0,100.0,2920222720.0,138935304192.0,339957,d040042bdb79aacdcafe80a2c4bd621271e3056823a84621c25ee8fd4fff594b
|
| 3 |
+
moe,sdnq,02_macro_pastry_glaze,4202,158.2769890979398,2.1681779328484905,36065.68798828125,40756.0,42504.0,100.0,3105587200.0,90820800512.0,367773,bcb2caabb9e29427019ef3dcc7790a89910c1b570e65f85bac3f8a95dc1912db
|
| 4 |
+
moe,sdnq,03_rain_market_lanterns,4203,158.72098272107542,2.1742600372750056,36067.5908203125,40756.0,41878.0,100.0,3175432192.0,91215613952.0,360438,e91a616e88278e062635d12bd204679caf879dafc107cb066cf4155c3efe2486
|
| 5 |
+
moe,sdnq,04_underwater_manta,4204,158.60668343293946,2.1726942936019102,36067.27880859375,40756.0,41878.0,100.0,3106979840.0,90012438528.0,359678,add7b1153f552850a2b095afa51998c4e3af4719f3d5c64425105f1af8c7458e
|
| 6 |
+
moe,sdnq,05_factory_battery_cells,4205,158.87487718299963,2.176368180589036,36067.6962890625,40756.0,41878.0,100.0,3307651072.0,87309594624.0,138406,7b6909a2aa49397f574338560cce4a3e47bb1930e55304f897523fa436fe7333
|
benchmark/sdnq.metrics.jsonl
ADDED
|
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"batch_cfg": false, "cold_generation": true, "ffprobe": {"programs": [], "streams": [{"codec_name": "h264", "duration": "3.041667", "height": 480, "nb_frames": "73", "pix_fmt": "yuv420p", "r_frame_rate": "24/1", "width": 832}]}, "fps": 24, "generation_seconds": 159.77318301890045, "guidance_scale": 3.0, "height": 480, "model_key": "moe", "negative_prompt": "{\"universal_negative\": {\"visual_quality\": [\"low quality\", \"worst quality\", \"blurry\", \"pixelated\", \"jpeg artifacts\", \"low resolution\", \"unstable color\", \"color flicker\", \"underexposed\", \"overexposed\", \"invisible subject\", \"subject hidden in darkness\"], \"artistic_style\": [\"painting\", \"illustration\", \"drawing\", \"cartoon\", \"3d render\", \"cgi\", \"sketch\", \"digital art\"], \"composition_and_content\": [\"text\", \"watermark\", \"signature\", \"logo\", \"subtitles\", \"pillarboxed\", \"side bars\", \"portrait image in landscape frame\"], \"temporal_and_motion_stability\": [\"flickering\", \"jittery\", \"motion blur\", \"temporal inconsistency\", \"warping\", \"morphing\", \"incoherent motion\", \"unnatural movement\", \"static object with sudden jump\", \"frame-to-frame inconsistency\"], \"material_and_structure\": [\"plastic-like glass\", \"unrealistic texture\", \"deformed bottle\", \"liquid freezing improperly\", \"distorted reflections\"]}}", "null_cond_clone_zero": false, "num_frames": 73, "num_inference_steps": 40, "nvidia_smi_after": {"memory.total": "183359", "memory.used": "41878", "name": "NVIDIA B200", "power.draw": "259.76", "temperature.gpu": "42", "timestamp": "2026/07/10 08:43:39.302", "utilization.gpu": "0"}, "output_file": "results/benchmark/moe/sdnq/videos/01_arctic_research_drone.mp4", "output_sha256": "d040042bdb79aacdcafe80a2c4bd621271e3056823a84621c25ee8fd4fff594b", "output_size_bytes": 339957, "peak_disk_used_bytes": 188465917952.0, "peak_gpu_memory_used_mb": 41878.0, "peak_gpu_power_watts": 981.74, "peak_gpu_utilization_percent": 100.0, "peak_process_rss_bytes": 2920222720.0, "peak_system_used_bytes": 138935304192.0, "prompt": "{\"comprehensive_description\":{\"scene_content_description\":\"A cinematic dusk shot of an Arctic research outpost on a wind-polished ice shelf under green aurora ribbons. A compact orange rover with a roof-mounted lidar mast drives slowly between blue instrument crates and slender weather antennas. Fine snow dust trails behind the tires, and warm light glows from the station windows against the cold cyan landscape. The atmosphere is realistic, high-detail documentary science footage with crisp snow texture, vapor in the air, and physically plausible shadows.\",\"camera_movement_description\":\"The camera begins as a low, smooth drone tracking shot behind the rover, then arcs gently left to reveal the glowing station and aurora-filled sky while keeping the rover dominant in frame.\"},\"camera_info\":{\"color\":\"Cold cyan with warm orange accents\",\"frame_size\":\"Wide\",\"shot_type_angle\":\"Low angle tracking shot\",\"lens_size\":\"Wide lens\",\"composition\":\"Rover in lower center, research station in the background, aurora above\",\"lighting\":\"Soft twilight with window glow\",\"lighting_type\":\"Natural dusk and practical interior lights\"},\"world_knowledge\":[],\"prominent_elements\":[{\"name\":\"orange autonomous research rover\",\"description\":\"A rugged six-wheeled scientific rover with orange body panels, black tires, a lidar mast, and small blinking status lights.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.7s]\",\"action\":\"drives forward slowly across packed snow, tires compressing the surface\"},{\"timestamp\":\"[0.7s - 1.5s]\",\"action\":\"turns slightly left as the lidar mast rotates\"},{\"timestamp\":\"[1.5s - 2.0s]\",\"action\":\"continues toward the lit station while fine snow trails behind\"}],\"location\":\"lower center of the frame\",\"relative_size\":\"dominant\",\"shape_and_color\":\"low rectangular orange body with black wheels and grey sensor mast\",\"texture\":\"matte painted metal, rubber tires, frost on edges\",\"appearance_details\":\"roof lidar, small antennas, narrow headlights, compact cargo rack\",\"relationship\":\"the main moving subject, traveling toward the research outpost\",\"orientation\":\"moving away from the camera and slightly left\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"Arctic research station\",\"description\":\"A modular polar station made of connected white containers with warm yellow light in the windows.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"remains stationary as wind-blown snow passes in front of it\"}],\"location\":\"middle background\",\"relative_size\":\"large\",\"shape_and_color\":\"rectangular white modules with dark seams and glowing windows\",\"texture\":\"frosted metal panels and glass\",\"appearance_details\":\"small stairs, railings, antennas, blue equipment crates nearby\",\"relationship\":\"destination of the rover\",\"orientation\":\"angled three-quarter view\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"aurora ribbons\",\"description\":\"Green aurora curtains stretching across the darkening polar sky.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"shimmer slowly and drift laterally across the sky\"}],\"location\":\"upper half of the frame\",\"relative_size\":\"large\",\"shape_and_color\":\"soft green luminous ribbons\",\"texture\":\"translucent atmospheric glow\",\"appearance_details\":\"layered folds and faint stars behind\",\"relationship\":\"dominates the sky above the station\",\"orientation\":\"horizontal sweeping arcs\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":true,\"number_of_objects\":\"many luminous bands\"}]}", "prompt_file": "01_arctic_research_drone.json", "prompt_id": "01_arctic_research_drone", "raw_frames_file": "results/benchmark/moe/sdnq/raw_frames/01_arctic_research_drone.npy", "raw_frames_sha256": "d8c8386e9e015500d0766968a355de2d172b50c2308b2438978d9dc270f815fe", "raw_frames_size_bytes": 87459968, "sample_count": 553, "sample_interval_seconds": 0.25, "seconds_per_frame": 2.188673739984938, "seed": 4201, "shift": 3.0, "torch_allocated_after_mb": 27094.3193359375, "torch_peak_allocated_mb": 36067.29833984375, "torch_peak_reserved_mb": 40756.0, "torch_reserved_after_mb": 40756.0, "variant": "sdnq", "width": 832}
|
| 2 |
+
{"batch_cfg": false, "cold_generation": false, "ffprobe": {"programs": [], "streams": [{"codec_name": "h264", "duration": "3.041667", "height": 480, "nb_frames": "73", "pix_fmt": "yuv420p", "r_frame_rate": "24/1", "width": 832}]}, "fps": 24, "generation_seconds": 158.2769890979398, "guidance_scale": 3.0, "height": 480, "model_key": "moe", "negative_prompt": "{\"universal_negative\": {\"visual_quality\": [\"low quality\", \"worst quality\", \"blurry\", \"pixelated\", \"jpeg artifacts\", \"low resolution\", \"unstable color\", \"color flicker\", \"underexposed\", \"overexposed\", \"invisible subject\", \"subject hidden in darkness\"], \"artistic_style\": [\"painting\", \"illustration\", \"drawing\", \"cartoon\", \"3d render\", \"cgi\", \"sketch\", \"digital art\"], \"composition_and_content\": [\"text\", \"watermark\", \"signature\", \"logo\", \"subtitles\", \"pillarboxed\", \"side bars\", \"portrait image in landscape frame\"], \"temporal_and_motion_stability\": [\"flickering\", \"jittery\", \"motion blur\", \"temporal inconsistency\", \"warping\", \"morphing\", \"incoherent motion\", \"unnatural movement\", \"static object with sudden jump\", \"frame-to-frame inconsistency\"], \"material_and_structure\": [\"plastic-like glass\", \"unrealistic texture\", \"deformed bottle\", \"liquid freezing improperly\", \"distorted reflections\"]}}", "null_cond_clone_zero": false, "num_frames": 73, "num_inference_steps": 40, "nvidia_smi_after": {"memory.total": "183359", "memory.used": "41878", "name": "NVIDIA B200", "power.draw": "260.53", "temperature.gpu": "43", "timestamp": "2026/07/10 08:46:20.584", "utilization.gpu": "0"}, "output_file": "results/benchmark/moe/sdnq/videos/02_macro_pastry_glaze.mp4", "output_sha256": "bcb2caabb9e29427019ef3dcc7790a89910c1b570e65f85bac3f8a95dc1912db", "output_size_bytes": 367773, "peak_disk_used_bytes": 188554317824.0, "peak_gpu_memory_used_mb": 42504.0, "peak_gpu_power_watts": 976.17, "peak_gpu_utilization_percent": 100.0, "peak_process_rss_bytes": 3105587200.0, "peak_system_used_bytes": 90820800512.0, "prompt": "{\"comprehensive_description\":{\"scene_content_description\":\"An extreme macro food cinematography shot in a professional pastry kitchen. A glossy dark chocolate dome dessert sits on a rotating metal turntable. A thin stream of amber caramel glaze pours from above, flowing over the curved surface in slow ribbons and revealing tiny reflections from softbox lights. Crushed pistachio dust and edible gold flakes sit around the base on a matte black plate. The look is photorealistic, tactile, high-speed macro video with shallow depth of field and no text.\",\"camera_movement_description\":\"The camera holds a close macro angle with a subtle push-in while the dessert rotates clockwise on the turntable.\"},\"camera_info\":{\"color\":\"Deep brown, amber, green pistachio, gold highlights\",\"frame_size\":\"Extreme close up\",\"shot_type_angle\":\"Slight high angle\",\"lens_size\":\"Macro lens\",\"composition\":\"Dessert centered with glaze entering from top\",\"lighting\":\"Large softbox reflections\",\"lighting_type\":\"Controlled studio kitchen lighting\"},\"world_knowledge\":[],\"prominent_elements\":[{\"name\":\"chocolate dome dessert\",\"description\":\"A small mirror-glazed chocolate dome on a black plate, surrounded by pistachio dust and gold flakes.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"rotates slowly clockwise on a metal turntable\"}],\"location\":\"center of the frame\",\"relative_size\":\"dominant\",\"shape_and_color\":\"hemispherical dark chocolate dome on round black plate\",\"texture\":\"mirror-gloss surface with smooth curved reflections\",\"appearance_details\":\"tiny gold flakes, green pistachio crumbs, clean pastry plating\",\"relationship\":\"receives the caramel glaze\",\"orientation\":\"upright on the turntable\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"caramel glaze stream\",\"description\":\"A viscous amber caramel stream pouring from above and spreading over the dessert.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.8s]\",\"action\":\"falls as a narrow glossy stream onto the crown of the dome\"},{\"timestamp\":\"[0.8s - 1.6s]\",\"action\":\"forms slow ribbons that slide down the curved chocolate surface\"},{\"timestamp\":\"[1.6s - 2.0s]\",\"action\":\"collects at the base in a thin shiny ring\"}],\"location\":\"entering from top center and flowing over the dessert\",\"relative_size\":\"medium\",\"shape_and_color\":\"thin amber liquid ribbons\",\"texture\":\"viscous, glossy, translucent caramel\",\"appearance_details\":\"bright highlights and tiny bubbles\",\"relationship\":\"coats the chocolate dome\",\"orientation\":\"vertical stream becoming downward trails\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"}]}", "prompt_file": "02_macro_pastry_glaze.json", "prompt_id": "02_macro_pastry_glaze", "raw_frames_file": "results/benchmark/moe/sdnq/raw_frames/02_macro_pastry_glaze.npy", "raw_frames_sha256": "0c455e53a36b1ca3a8e933229529ff2541faf889e504d4527ad9795e8083ae6e", "raw_frames_size_bytes": 87459968, "sample_count": 538, "sample_interval_seconds": 0.25, "seconds_per_frame": 2.1681779328484905, "seed": 4202, "shift": 3.0, "torch_allocated_after_mb": 27094.3193359375, "torch_peak_allocated_mb": 36065.68798828125, "torch_peak_reserved_mb": 40756.0, "torch_reserved_after_mb": 40756.0, "variant": "sdnq", "width": 832}
|
| 3 |
+
{"batch_cfg": false, "cold_generation": false, "ffprobe": {"programs": [], "streams": [{"codec_name": "h264", "duration": "3.041667", "height": 480, "nb_frames": "73", "pix_fmt": "yuv420p", "r_frame_rate": "24/1", "width": 832}]}, "fps": 24, "generation_seconds": 158.72098272107542, "guidance_scale": 3.0, "height": 480, "model_key": "moe", "negative_prompt": "{\"universal_negative\": {\"visual_quality\": [\"low quality\", \"worst quality\", \"blurry\", \"pixelated\", \"jpeg artifacts\", \"low resolution\", \"unstable color\", \"color flicker\", \"underexposed\", \"overexposed\", \"invisible subject\", \"subject hidden in darkness\"], \"artistic_style\": [\"painting\", \"illustration\", \"drawing\", \"cartoon\", \"3d render\", \"cgi\", \"sketch\", \"digital art\"], \"composition_and_content\": [\"text\", \"watermark\", \"signature\", \"logo\", \"subtitles\", \"pillarboxed\", \"side bars\", \"portrait image in landscape frame\"], \"temporal_and_motion_stability\": [\"flickering\", \"jittery\", \"motion blur\", \"temporal inconsistency\", \"warping\", \"morphing\", \"incoherent motion\", \"unnatural movement\", \"static object with sudden jump\", \"frame-to-frame inconsistency\"], \"material_and_structure\": [\"plastic-like glass\", \"unrealistic texture\", \"deformed bottle\", \"liquid freezing improperly\", \"distorted reflections\"]}}", "null_cond_clone_zero": false, "num_frames": 73, "num_inference_steps": 40, "nvidia_smi_after": {"memory.total": "183359", "memory.used": "41878", "name": "NVIDIA B200", "power.draw": "260.26", "temperature.gpu": "43", "timestamp": "2026/07/10 08:49:02.315", "utilization.gpu": "0"}, "output_file": "results/benchmark/moe/sdnq/videos/03_rain_market_lanterns.mp4", "output_sha256": "e91a616e88278e062635d12bd204679caf879dafc107cb066cf4155c3efe2486", "output_size_bytes": 360438, "peak_disk_used_bytes": 183805046784.0, "peak_gpu_memory_used_mb": 41878.0, "peak_gpu_power_watts": 975.4, "peak_gpu_utilization_percent": 100.0, "peak_process_rss_bytes": 3175432192.0, "peak_system_used_bytes": 91215613952.0, "prompt": "{\"comprehensive_description\":{\"scene_content_description\":\"A lively rainy night market in a narrow old-city street, filmed like naturalistic cinema. Red paper lanterns hang from wooden stalls, steam rises from bowls of soup, and wet cobblestones mirror orange and teal reflections. A vendor in a dark apron slides a steaming bowl across a counter to a traveler in a blue raincoat. Umbrellas pass in the foreground, partially occluding the scene for a moment. The mood is warm, humid, detailed, and grounded, with no readable signs or subtitles.\",\"camera_movement_description\":\"The camera performs a slow lateral dolly from left to right at counter height, with foreground umbrellas briefly crossing the lens.\"},\"camera_info\":{\"color\":\"Warm lantern orange, teal rain reflections, red accents\",\"frame_size\":\"Medium wide\",\"shot_type_angle\":\"Eye level\",\"lens_size\":\"Medium lens\",\"composition\":\"Vendor stall centered, traveler on right, passing umbrellas in foreground\",\"lighting\":\"Lantern light mixed with rain reflections\",\"lighting_type\":\"Night practical lighting\"},\"world_knowledge\":[],\"prominent_elements\":[{\"name\":\"food vendor\",\"description\":\"A middle-aged street vendor wearing a dark apron and working behind a steaming wooden food stall.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.8s]\",\"action\":\"lifts a white ceramic bowl from a steaming pot\"},{\"timestamp\":\"[0.8s - 1.6s]\",\"action\":\"slides the bowl carefully across the counter\"},{\"timestamp\":\"[1.6s - 2.0s]\",\"action\":\"wipes condensation from the counter with a cloth\"}],\"location\":\"center-left behind the stall counter\",\"relative_size\":\"large\",\"shape_and_color\":\"human figure in dark apron under warm lantern light\",\"texture\":\"wet fabric, skin highlights, wood grain around the stall\",\"appearance_details\":\"rolled sleeves, focused expression, steam around hands\",\"relationship\":\"serves food to the traveler\",\"orientation\":\"facing right toward the customer\",\"pose\":\"leaning slightly forward over the counter\",\"expression\":\"focused and calm\",\"clothing\":\"dark apron over a muted shirt\",\"gender\":\"\",\"skin_tone_and_texture\":\"natural skin with rain-lit highlights\"},{\"name\":\"traveler in blue raincoat\",\"description\":\"A traveler wearing a blue hooded raincoat, waiting at the stall with wet sleeves and a folded umbrella.\",\"actions\":[{\"timestamp\":\"[0.0s - 1.2s]\",\"action\":\"waits with hands near the counter, watching the bowl\"},{\"timestamp\":\"[1.2s - 2.0s]\",\"action\":\"reaches forward to receive the steaming bowl\"}],\"location\":\"right side of the frame\",\"relative_size\":\"medium\",\"shape_and_color\":\"blue hooded raincoat silhouette\",\"texture\":\"water beads on waterproof fabric\",\"appearance_details\":\"hood up, folded umbrella tucked under one arm\",\"relationship\":\"customer receiving food from the vendor\",\"orientation\":\"facing left toward the stall\",\"pose\":\"standing close to the counter\",\"expression\":\"expectant and tired\",\"clothing\":\"blue raincoat\",\"gender\":\"\",\"skin_tone_and_texture\":\"\"},{\"name\":\"rain and lantern reflections\",\"description\":\"Falling rain, steam, and glowing reflections on wet cobblestones.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"rain falls continuously while reflections ripple on the street\"}],\"location\":\"foreground and background throughout the frame\",\"relative_size\":\"large\",\"shape_and_color\":\"thin rain streaks, orange and teal reflected pools\",\"texture\":\"wet stone, vapor, glossy water\",\"appearance_details\":\"soft steam plumes, umbrellas passing close to lens\",\"relationship\":\"sets the atmosphere around the market\",\"orientation\":\"vertical rainfall and horizontal street reflections\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":true,\"number_of_objects\":\"many raindrops and reflections\"}]}", "prompt_file": "03_rain_market_lanterns.json", "prompt_id": "03_rain_market_lanterns", "raw_frames_file": "results/benchmark/moe/sdnq/raw_frames/03_rain_market_lanterns.npy", "raw_frames_sha256": "535b3669bd1fc07990872df951b2544288ac4e4554000d9c4fb705a24dd33ed3", "raw_frames_size_bytes": 87459968, "sample_count": 549, "sample_interval_seconds": 0.25, "seconds_per_frame": 2.1742600372750056, "seed": 4203, "shift": 3.0, "torch_allocated_after_mb": 27094.3193359375, "torch_peak_allocated_mb": 36067.5908203125, "torch_peak_reserved_mb": 40756.0, "torch_reserved_after_mb": 40756.0, "variant": "sdnq", "width": 832}
|
| 4 |
+
{"batch_cfg": false, "cold_generation": false, "ffprobe": {"programs": [], "streams": [{"codec_name": "h264", "duration": "3.041667", "height": 480, "nb_frames": "73", "pix_fmt": "yuv420p", "r_frame_rate": "24/1", "width": 832}]}, "fps": 24, "generation_seconds": 158.60668343293946, "guidance_scale": 3.0, "height": 480, "model_key": "moe", "negative_prompt": "{\"universal_negative\": {\"visual_quality\": [\"low quality\", \"worst quality\", \"blurry\", \"pixelated\", \"jpeg artifacts\", \"low resolution\", \"unstable color\", \"color flicker\", \"underexposed\", \"overexposed\", \"invisible subject\", \"subject hidden in darkness\"], \"artistic_style\": [\"painting\", \"illustration\", \"drawing\", \"cartoon\", \"3d render\", \"cgi\", \"sketch\", \"digital art\"], \"composition_and_content\": [\"text\", \"watermark\", \"signature\", \"logo\", \"subtitles\", \"pillarboxed\", \"side bars\", \"portrait image in landscape frame\"], \"temporal_and_motion_stability\": [\"flickering\", \"jittery\", \"motion blur\", \"temporal inconsistency\", \"warping\", \"morphing\", \"incoherent motion\", \"unnatural movement\", \"static object with sudden jump\", \"frame-to-frame inconsistency\"], \"material_and_structure\": [\"plastic-like glass\", \"unrealistic texture\", \"deformed bottle\", \"liquid freezing improperly\", \"distorted reflections\"]}}", "null_cond_clone_zero": false, "num_frames": 73, "num_inference_steps": 40, "nvidia_smi_after": {"memory.total": "183359", "memory.used": "41878", "name": "NVIDIA B200", "power.draw": "260.36", "temperature.gpu": "43", "timestamp": "2026/07/10 08:51:43.973", "utilization.gpu": "0"}, "output_file": "results/benchmark/moe/sdnq/videos/04_underwater_manta.mp4", "output_sha256": "add7b1153f552850a2b095afa51998c4e3af4719f3d5c64425105f1af8c7458e", "output_size_bytes": 359678, "peak_disk_used_bytes": 183893323776.0, "peak_gpu_memory_used_mb": 41878.0, "peak_gpu_power_watts": 981.58, "peak_gpu_utilization_percent": 100.0, "peak_process_rss_bytes": 3106979840.0, "peak_system_used_bytes": 90012438528.0, "prompt": "{\"comprehensive_description\":{\"scene_content_description\":\"A tranquil underwater wildlife scene in clear tropical water. A large manta ray glides above a colorful coral garden while a scuba diver hovers several meters behind, holding a small camera rig. Sunbeams ripple through the surface, schools of tiny silver fish split and rejoin around the manta, and suspended particles drift slowly. The image should feel like realistic underwater documentary footage with natural motion, soft blue-green color, and no fantasy elements.\",\"camera_movement_description\":\"The camera tracks smoothly beside the manta ray from a slightly lower angle, drifting forward with the current.\"},\"camera_info\":{\"color\":\"Blue-green water with coral reds and yellows\",\"frame_size\":\"Wide\",\"shot_type_angle\":\"Slight low angle\",\"lens_size\":\"Wide underwater lens\",\"composition\":\"Manta ray crossing the upper center, coral below, diver in background\",\"lighting\":\"Dappled sunlight\",\"lighting_type\":\"Natural underwater daylight\"},\"world_knowledge\":[],\"prominent_elements\":[{\"name\":\"manta ray\",\"description\":\"A large manta ray with broad triangular fins, dark top surface, and pale underside.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.8s]\",\"action\":\"glides from left to right with slow wing-like fin movement\"},{\"timestamp\":\"[0.8s - 1.5s]\",\"action\":\"tilts slightly upward as sunbeams cross its back\"},{\"timestamp\":\"[1.5s - 2.0s]\",\"action\":\"continues forward while small fish scatter around it\"}],\"location\":\"upper center moving toward right\",\"relative_size\":\"dominant\",\"shape_and_color\":\"wide diamond-like silhouette, dark grey top, pale underside\",\"texture\":\"smooth skin with subtle mottling\",\"appearance_details\":\"cephalic fins near mouth, long tail trailing behind\",\"relationship\":\"main wildlife subject of the video\",\"orientation\":\"moving left to right and slightly upward\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"scuba diver\",\"description\":\"A diver in black wetsuit and fins, carrying a small underwater camera rig.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"hovers calmly in the background, exhaling small bubble streams\"}],\"location\":\"mid-background behind the manta ray\",\"relative_size\":\"small\",\"shape_and_color\":\"human figure in black wetsuit with silver air tank\",\"texture\":\"neoprene suit, metal tank, glass mask\",\"appearance_details\":\"fins, mask, regulator, compact camera lights\",\"relationship\":\"observes the manta ray without touching it\",\"orientation\":\"facing the manta ray\",\"pose\":\"horizontal hover\",\"expression\":\"\",\"clothing\":\"black scuba gear\",\"gender\":\"\",\"skin_tone_and_texture\":\"\"},{\"name\":\"coral reef and fish\",\"description\":\"A colorful coral garden with small silver fish schooling above it.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"fish swirl and split around the manta while coral remains fixed\"}],\"location\":\"lower half of the frame\",\"relative_size\":\"large\",\"shape_and_color\":\"branching corals in red, yellow, and muted purple; silver fish clusters\",\"texture\":\"rough coral, shimmering fish scales\",\"appearance_details\":\"sand patches, sea fans, drifting particles\",\"relationship\":\"environment below the manta ray\",\"orientation\":\"reef spreads horizontally across the bottom\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":true,\"number_of_objects\":\"many fish and coral structures\"}]}", "prompt_file": "04_underwater_manta.json", "prompt_id": "04_underwater_manta", "raw_frames_file": "results/benchmark/moe/sdnq/raw_frames/04_underwater_manta.npy", "raw_frames_sha256": "7a51833e767e2dffec08ef8a47ab57a5541ad34cdedc30553362369cd287f2c9", "raw_frames_size_bytes": 87459968, "sample_count": 555, "sample_interval_seconds": 0.25, "seconds_per_frame": 2.1726942936019102, "seed": 4204, "shift": 3.0, "torch_allocated_after_mb": 27094.3193359375, "torch_peak_allocated_mb": 36067.27880859375, "torch_peak_reserved_mb": 40756.0, "torch_reserved_after_mb": 40756.0, "variant": "sdnq", "width": 832}
|
| 5 |
+
{"batch_cfg": false, "cold_generation": false, "ffprobe": {"programs": [], "streams": [{"codec_name": "h264", "duration": "3.041667", "height": 480, "nb_frames": "73", "pix_fmt": "yuv420p", "r_frame_rate": "24/1", "width": 832}]}, "fps": 24, "generation_seconds": 158.87487718299963, "guidance_scale": 3.0, "height": 480, "model_key": "moe", "negative_prompt": "{\"universal_negative\": {\"visual_quality\": [\"low quality\", \"worst quality\", \"blurry\", \"pixelated\", \"jpeg artifacts\", \"low resolution\", \"unstable color\", \"color flicker\", \"underexposed\", \"overexposed\", \"invisible subject\", \"subject hidden in darkness\"], \"artistic_style\": [\"painting\", \"illustration\", \"drawing\", \"cartoon\", \"3d render\", \"cgi\", \"sketch\", \"digital art\"], \"composition_and_content\": [\"text\", \"watermark\", \"signature\", \"logo\", \"subtitles\", \"pillarboxed\", \"side bars\", \"portrait image in landscape frame\"], \"temporal_and_motion_stability\": [\"flickering\", \"jittery\", \"motion blur\", \"temporal inconsistency\", \"warping\", \"morphing\", \"incoherent motion\", \"unnatural movement\", \"static object with sudden jump\", \"frame-to-frame inconsistency\"], \"material_and_structure\": [\"plastic-like glass\", \"unrealistic texture\", \"deformed bottle\", \"liquid freezing improperly\", \"distorted reflections\"]}}", "null_cond_clone_zero": false, "num_frames": 73, "num_inference_steps": 40, "nvidia_smi_after": {"memory.total": "183359", "memory.used": "41878", "name": "NVIDIA B200", "power.draw": "260.35", "temperature.gpu": "43", "timestamp": "2026/07/10 08:54:25.766", "utilization.gpu": "0"}, "output_file": "results/benchmark/moe/sdnq/videos/05_factory_battery_cells.mp4", "output_sha256": "7b6909a2aa49397f574338560cce4a3e47bb1930e55304f897523fa436fe7333", "output_size_bytes": 138406, "peak_disk_used_bytes": 183981420544.0, "peak_gpu_memory_used_mb": 41878.0, "peak_gpu_power_watts": 981.13, "peak_gpu_utilization_percent": 100.0, "peak_process_rss_bytes": 3307651072.0, "peak_system_used_bytes": 87309594624.0, "prompt": "{\"comprehensive_description\":{\"scene_content_description\":\"A clean high-tech factory line assembling transparent solid-state battery cells. Two white robotic arms move with precise synchronized motion over a brushed steel conveyor. One arm lowers a translucent rectangular cell into a copper test fixture while the other arm scans it with a blue light bar. Tiny reflections slide across glass safety panels, and status lights pulse softly. The scene should look like realistic industrial automation footage, not CGI, with crisp metal surfaces and controlled motion.\",\"camera_movement_description\":\"The camera is locked off in a medium-wide three-quarter view with a subtle mechanical vibration, emphasizing repeatable robotic motion.\"},\"camera_info\":{\"color\":\"White, brushed steel, copper, blue inspection light\",\"frame_size\":\"Medium wide\",\"shot_type_angle\":\"Three-quarter eye-level industrial view\",\"lens_size\":\"Medium lens\",\"composition\":\"Two robotic arms framing the battery cell at center\",\"lighting\":\"Even overhead factory lighting with blue scanner highlight\",\"lighting_type\":\"Artificial industrial lighting\"},\"world_knowledge\":[],\"prominent_elements\":[{\"name\":\"left robotic arm\",\"description\":\"A white six-axis robotic arm with a vacuum gripper holding a transparent rectangular battery cell.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.7s]\",\"action\":\"moves downward from upper left with the transparent cell held steady\"},{\"timestamp\":\"[0.7s - 1.4s]\",\"action\":\"places the cell into a copper test fixture\"},{\"timestamp\":\"[1.4s - 2.0s]\",\"action\":\"releases the cell and retracts slightly upward\"}],\"location\":\"left side moving toward center\",\"relative_size\":\"large\",\"shape_and_color\":\"white articulated segments with black joints\",\"texture\":\"smooth painted metal and rubber vacuum cups\",\"appearance_details\":\"visible cable routing, small green status LED\",\"relationship\":\"positions the battery cell for testing\",\"orientation\":\"angled downward toward the center fixture\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"right robotic scanner arm\",\"description\":\"A second white robotic arm carrying a rectangular blue inspection light bar.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.9s]\",\"action\":\"waits above the copper fixture with blue scanner light dim\"},{\"timestamp\":\"[0.9s - 1.6s]\",\"action\":\"sweeps the blue light bar across the transparent cell\"},{\"timestamp\":\"[1.6s - 2.0s]\",\"action\":\"pauses as the scanner light pulses once\"}],\"location\":\"right side above the conveyor\",\"relative_size\":\"large\",\"shape_and_color\":\"white arm with glowing blue rectangular scanner\",\"texture\":\"smooth metal casing and glass scanner cover\",\"appearance_details\":\"blue light strip, black joints, compact sensor module\",\"relationship\":\"inspects the cell after placement\",\"orientation\":\"angled left toward the battery cell\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"transparent battery cell and copper fixture\",\"description\":\"A clear rectangular solid-state cell seated into a copper test fixture on a steel conveyor.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.7s]\",\"action\":\"moves with the left arm above the fixture\"},{\"timestamp\":\"[0.7s - 1.4s]\",\"action\":\"is lowered precisely into the fixture\"},{\"timestamp\":\"[1.4s - 2.0s]\",\"action\":\"remains stationary while blue inspection light passes over it\"}],\"location\":\"center of the frame on the conveyor\",\"relative_size\":\"medium\",\"shape_and_color\":\"transparent rectangle with faint internal layers, copper fixture\",\"texture\":\"clear glass-like cell, brushed copper, polished steel\",\"appearance_details\":\"thin internal laminate layers, small alignment pins, clean conveyor surface\",\"relationship\":\"object being assembled and inspected\",\"orientation\":\"horizontal in the fixture\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"}]}", "prompt_file": "05_factory_battery_cells.json", "prompt_id": "05_factory_battery_cells", "raw_frames_file": "results/benchmark/moe/sdnq/raw_frames/05_factory_battery_cells.npy", "raw_frames_sha256": "2a8b445ef18edfe1789e3d4b779b9a8eb290055605f24b4e7cef890f927a3e5e", "raw_frames_size_bytes": 87459968, "sample_count": 556, "sample_interval_seconds": 0.25, "seconds_per_frame": 2.176368180589036, "seed": 4205, "shift": 3.0, "torch_allocated_after_mb": 27094.3193359375, "torch_peak_allocated_mb": 36067.6962890625, "torch_peak_reserved_mb": 40756.0, "torch_reserved_after_mb": 40756.0, "variant": "sdnq", "width": 832}
|
benchmark/shared/01_arctic_research_drone.meta.json
ADDED
|
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"base_video": "results/benchmark/moe/original/videos/01_arctic_research_drone.mp4",
|
| 3 |
+
"base_video_sha256": "e84547fdbd70c1f118435c0b0dee74ccb719a64f1c51e2d7b1260f2185f3dad9",
|
| 4 |
+
"fps": 24,
|
| 5 |
+
"height": 1088,
|
| 6 |
+
"latent_dtype": "torch.float32",
|
| 7 |
+
"latent_sha256": "502b10f841d96aa101e69421b20083aeb60427c054b42fb0db29c5a6e70824cf",
|
| 8 |
+
"latent_shape": [
|
| 9 |
+
1,
|
| 10 |
+
16,
|
| 11 |
+
19,
|
| 12 |
+
136,
|
| 13 |
+
240
|
| 14 |
+
],
|
| 15 |
+
"lowres_meta": {
|
| 16 |
+
"align_to_training": true,
|
| 17 |
+
"max_frames": 81,
|
| 18 |
+
"num_source_frames": 73,
|
| 19 |
+
"sample_frame": 73,
|
| 20 |
+
"sample_frame_uncapped": 73,
|
| 21 |
+
"src_fps": 24.0,
|
| 22 |
+
"t_vae": 19,
|
| 23 |
+
"truncated_by_max_frames": false,
|
| 24 |
+
"vae_fps": 24.0
|
| 25 |
+
},
|
| 26 |
+
"seed": 4201,
|
| 27 |
+
"t_thresh": 0.85,
|
| 28 |
+
"width": 1920
|
| 29 |
+
}
|
benchmark/smokes/moe-refiner-sdnq-standard.json
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"finite": true,
|
| 3 |
+
"frame_shape": [
|
| 4 |
+
9,
|
| 5 |
+
192,
|
| 6 |
+
320,
|
| 7 |
+
3
|
| 8 |
+
],
|
| 9 |
+
"generation_seconds": 2.7460155420703813,
|
| 10 |
+
"load_seconds": 40.40503731998615,
|
| 11 |
+
"model": "releases/WaveCut-LingBot-Video-MoE-30B-A3B-SDNQ-uint4-static",
|
| 12 |
+
"offload_mode": "standard",
|
| 13 |
+
"output": "results/smoke/moe-refiner-sdnq-standard.mp4",
|
| 14 |
+
"torch_peak_allocated_mb": 29051.501953125,
|
| 15 |
+
"torch_peak_reserved_mb": 30770.0,
|
| 16 |
+
"transformer_subfolder": "refiner"
|
| 17 |
+
}
|
benchmark/smokes/moe-refiner-sdnq-standard.mp4
ADDED
|
Binary file (11.5 kB). View file
|
|
|
benchmark/smokes/moe-sdnq-model.json
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"finite": true,
|
| 3 |
+
"frame_shape": [
|
| 4 |
+
9,
|
| 5 |
+
192,
|
| 6 |
+
320,
|
| 7 |
+
3
|
| 8 |
+
],
|
| 9 |
+
"generation_seconds": 25.569705806905404,
|
| 10 |
+
"load_seconds": 37.127981438999996,
|
| 11 |
+
"model": "releases/WaveCut-LingBot-Video-MoE-30B-A3B-SDNQ-uint4-static",
|
| 12 |
+
"offload_mode": "model",
|
| 13 |
+
"output": "results/smoke/moe-sdnq-model.mp4",
|
| 14 |
+
"torch_peak_allocated_mb": 20024.828125,
|
| 15 |
+
"torch_peak_reserved_mb": 20666.0,
|
| 16 |
+
"transformer_subfolder": "transformer"
|
| 17 |
+
}
|
benchmark/smokes/moe-sdnq-model.mp4
ADDED
|
Binary file (14.5 kB). View file
|
|
|
benchmark/smokes/moe-sdnq-sequential.json
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"finite": true,
|
| 3 |
+
"frame_shape": [
|
| 4 |
+
9,
|
| 5 |
+
192,
|
| 6 |
+
320,
|
| 7 |
+
3
|
| 8 |
+
],
|
| 9 |
+
"generation_seconds": 22.574472793960012,
|
| 10 |
+
"load_seconds": 41.77268342196476,
|
| 11 |
+
"model": "releases/WaveCut-LingBot-Video-MoE-30B-A3B-SDNQ-uint4-static",
|
| 12 |
+
"offload_mode": "sequential",
|
| 13 |
+
"output": "results/smoke/moe-sdnq-sequential.mp4",
|
| 14 |
+
"torch_peak_allocated_mb": 2109.29296875,
|
| 15 |
+
"torch_peak_reserved_mb": 2712.0,
|
| 16 |
+
"transformer_subfolder": "transformer"
|
| 17 |
+
}
|
benchmark/smokes/moe-sdnq-sequential.mp4
ADDED
|
Binary file (14.5 kB). View file
|
|
|
benchmark/smokes/moe-sdnq-standard.json
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"finite": true,
|
| 3 |
+
"frame_shape": [
|
| 4 |
+
9,
|
| 5 |
+
192,
|
| 6 |
+
320,
|
| 7 |
+
3
|
| 8 |
+
],
|
| 9 |
+
"generation_seconds": 4.120272382977419,
|
| 10 |
+
"load_seconds": 41.03481481503695,
|
| 11 |
+
"model": "releases/WaveCut-LingBot-Video-MoE-30B-A3B-SDNQ-uint4-static",
|
| 12 |
+
"offload_mode": "standard",
|
| 13 |
+
"output": "results/smoke/moe-sdnq-standard.mp4",
|
| 14 |
+
"torch_peak_allocated_mb": 29051.501953125,
|
| 15 |
+
"torch_peak_reserved_mb": 30770.0,
|
| 16 |
+
"transformer_subfolder": "transformer"
|
| 17 |
+
}
|
benchmark/smokes/moe-sdnq-standard.mp4
ADDED
|
Binary file (14.5 kB). View file
|
|
|
benchmark/summary.json
ADDED
|
@@ -0,0 +1,1218 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"base": {
|
| 3 |
+
"original": {
|
| 4 |
+
"cold_generation_seconds": 140.9023737310199,
|
| 5 |
+
"environment": {
|
| 6 |
+
"cuda": "12.8",
|
| 7 |
+
"gpu": "NVIDIA B200",
|
| 8 |
+
"lingbot_moe_expert_backend": "sglang_triton",
|
| 9 |
+
"python": "3.12.3 (main, Aug 14 2025, 17:47:21) [GCC 13.3.0]",
|
| 10 |
+
"torch": "2.8.0+cu128"
|
| 11 |
+
},
|
| 12 |
+
"generation_mean_seconds": 139.886719101551,
|
| 13 |
+
"hot_generation_mean_seconds": 139.63280544418376,
|
| 14 |
+
"load": {
|
| 15 |
+
"nvidia_smi_after": {
|
| 16 |
+
"memory.total": "183359",
|
| 17 |
+
"memory.used": "67234",
|
| 18 |
+
"name": "NVIDIA B200",
|
| 19 |
+
"power.draw": "249.62",
|
| 20 |
+
"temperature.gpu": "31",
|
| 21 |
+
"timestamp": "2026/07/10 08:27:42.682",
|
| 22 |
+
"utilization.gpu": "14"
|
| 23 |
+
},
|
| 24 |
+
"peak_disk_used_bytes": 188018044928.0,
|
| 25 |
+
"peak_gpu_memory_used_mb": 66692.0,
|
| 26 |
+
"peak_gpu_power_watts": 249.71,
|
| 27 |
+
"peak_gpu_utilization_percent": 19.0,
|
| 28 |
+
"peak_process_rss_bytes": 10081595392.0,
|
| 29 |
+
"peak_system_used_bytes": 78059479040.0,
|
| 30 |
+
"sample_count": 154,
|
| 31 |
+
"sample_interval_seconds": 0.25,
|
| 32 |
+
"seconds": 50.28812898299657,
|
| 33 |
+
"torch_allocated_after_mb": 66603.46142578125,
|
| 34 |
+
"torch_peak_allocated_mb": 66603.46142578125,
|
| 35 |
+
"torch_peak_reserved_mb": 66608.0,
|
| 36 |
+
"torch_reserved_after_mb": 66608.0
|
| 37 |
+
},
|
| 38 |
+
"outputs": [
|
| 39 |
+
{
|
| 40 |
+
"ffprobe": {
|
| 41 |
+
"programs": [],
|
| 42 |
+
"streams": [
|
| 43 |
+
{
|
| 44 |
+
"codec_name": "h264",
|
| 45 |
+
"duration": "3.041667",
|
| 46 |
+
"height": 480,
|
| 47 |
+
"nb_frames": "73",
|
| 48 |
+
"pix_fmt": "yuv420p",
|
| 49 |
+
"r_frame_rate": "24/1",
|
| 50 |
+
"width": 832
|
| 51 |
+
}
|
| 52 |
+
]
|
| 53 |
+
},
|
| 54 |
+
"generation_seconds": 140.9023737310199,
|
| 55 |
+
"prompt_id": "01_arctic_research_drone",
|
| 56 |
+
"seed": 4201,
|
| 57 |
+
"sha256": "7a548d09dd53755c1fdd7e759aae7cf92fd57d555bdc13f28023ede4ffe5f95e",
|
| 58 |
+
"size_bytes": 366044
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"ffprobe": {
|
| 62 |
+
"programs": [],
|
| 63 |
+
"streams": [
|
| 64 |
+
{
|
| 65 |
+
"codec_name": "h264",
|
| 66 |
+
"duration": "3.041667",
|
| 67 |
+
"height": 480,
|
| 68 |
+
"nb_frames": "73",
|
| 69 |
+
"pix_fmt": "yuv420p",
|
| 70 |
+
"r_frame_rate": "24/1",
|
| 71 |
+
"width": 832
|
| 72 |
+
}
|
| 73 |
+
]
|
| 74 |
+
},
|
| 75 |
+
"generation_seconds": 139.39877224492375,
|
| 76 |
+
"prompt_id": "02_macro_pastry_glaze",
|
| 77 |
+
"seed": 4202,
|
| 78 |
+
"sha256": "0073c4f94ed8fc0b9bfaa93966ac627b8a66f31efe9547a78bc6934ee2ecfd82",
|
| 79 |
+
"size_bytes": 396167
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"ffprobe": {
|
| 83 |
+
"programs": [],
|
| 84 |
+
"streams": [
|
| 85 |
+
{
|
| 86 |
+
"codec_name": "h264",
|
| 87 |
+
"duration": "3.041667",
|
| 88 |
+
"height": 480,
|
| 89 |
+
"nb_frames": "73",
|
| 90 |
+
"pix_fmt": "yuv420p",
|
| 91 |
+
"r_frame_rate": "24/1",
|
| 92 |
+
"width": 832
|
| 93 |
+
}
|
| 94 |
+
]
|
| 95 |
+
},
|
| 96 |
+
"generation_seconds": 139.6812952979235,
|
| 97 |
+
"prompt_id": "03_rain_market_lanterns",
|
| 98 |
+
"seed": 4203,
|
| 99 |
+
"sha256": "753f2016262fbd6a3644b28ad4f6243e15e8b55ef9474616a9c7d636f474090b",
|
| 100 |
+
"size_bytes": 430040
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"ffprobe": {
|
| 104 |
+
"programs": [],
|
| 105 |
+
"streams": [
|
| 106 |
+
{
|
| 107 |
+
"codec_name": "h264",
|
| 108 |
+
"duration": "3.041667",
|
| 109 |
+
"height": 480,
|
| 110 |
+
"nb_frames": "73",
|
| 111 |
+
"pix_fmt": "yuv420p",
|
| 112 |
+
"r_frame_rate": "24/1",
|
| 113 |
+
"width": 832
|
| 114 |
+
}
|
| 115 |
+
]
|
| 116 |
+
},
|
| 117 |
+
"generation_seconds": 139.6351271129679,
|
| 118 |
+
"prompt_id": "04_underwater_manta",
|
| 119 |
+
"seed": 4204,
|
| 120 |
+
"sha256": "dcd3ffe1e7dd1f72f86e256d0db51acfb6661b880bfb0b2a5b798c9609cae24a",
|
| 121 |
+
"size_bytes": 359114
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"ffprobe": {
|
| 125 |
+
"programs": [],
|
| 126 |
+
"streams": [
|
| 127 |
+
{
|
| 128 |
+
"codec_name": "h264",
|
| 129 |
+
"duration": "3.041667",
|
| 130 |
+
"height": 480,
|
| 131 |
+
"nb_frames": "73",
|
| 132 |
+
"pix_fmt": "yuv420p",
|
| 133 |
+
"r_frame_rate": "24/1",
|
| 134 |
+
"width": 832
|
| 135 |
+
}
|
| 136 |
+
]
|
| 137 |
+
},
|
| 138 |
+
"generation_seconds": 139.8160271209199,
|
| 139 |
+
"prompt_id": "05_factory_battery_cells",
|
| 140 |
+
"seed": 4205,
|
| 141 |
+
"sha256": "8a08377ecb3877e3ab261ca1840450dfa28a18ccc6cc59679002da1ff7750b49",
|
| 142 |
+
"size_bytes": 324116
|
| 143 |
+
}
|
| 144 |
+
],
|
| 145 |
+
"peak_gpu_memory_used_mb": 126274.0,
|
| 146 |
+
"peak_process_rss_bytes": 3218972672.0,
|
| 147 |
+
"peak_system_used_bytes": 92903628800.0,
|
| 148 |
+
"peak_torch_allocated_mb": 112470.375,
|
| 149 |
+
"settings": {
|
| 150 |
+
"batch_cfg": false,
|
| 151 |
+
"fps": 24,
|
| 152 |
+
"guidance_scale": 3.0,
|
| 153 |
+
"height": 480,
|
| 154 |
+
"null_cond_clone_zero": false,
|
| 155 |
+
"num_frames": 73,
|
| 156 |
+
"num_inference_steps": 40,
|
| 157 |
+
"shift": 3.0,
|
| 158 |
+
"width": 832
|
| 159 |
+
},
|
| 160 |
+
"variant": "original"
|
| 161 |
+
},
|
| 162 |
+
"sdnq": {
|
| 163 |
+
"cold_generation_seconds": 159.77318301890045,
|
| 164 |
+
"environment": {
|
| 165 |
+
"cuda": "12.8",
|
| 166 |
+
"gpu": "NVIDIA B200",
|
| 167 |
+
"lingbot_moe_expert_backend": "sglang_triton",
|
| 168 |
+
"python": "3.12.3 (main, Aug 14 2025, 17:47:21) [GCC 13.3.0]",
|
| 169 |
+
"torch": "2.8.0+cu128"
|
| 170 |
+
},
|
| 171 |
+
"generation_mean_seconds": 158.85054309077094,
|
| 172 |
+
"hot_generation_mean_seconds": 158.61988310873858,
|
| 173 |
+
"load": {
|
| 174 |
+
"nvidia_smi_after": {
|
| 175 |
+
"memory.total": "183359",
|
| 176 |
+
"memory.used": "27698",
|
| 177 |
+
"name": "NVIDIA B200",
|
| 178 |
+
"power.draw": "250.43",
|
| 179 |
+
"temperature.gpu": "32",
|
| 180 |
+
"timestamp": "2026/07/10 08:40:56.535",
|
| 181 |
+
"utilization.gpu": "17"
|
| 182 |
+
},
|
| 183 |
+
"peak_disk_used_bytes": 188465856512.0,
|
| 184 |
+
"peak_gpu_memory_used_mb": 27188.0,
|
| 185 |
+
"peak_gpu_power_watts": 250.52,
|
| 186 |
+
"peak_gpu_utilization_percent": 19.0,
|
| 187 |
+
"peak_process_rss_bytes": 10137681920.0,
|
| 188 |
+
"peak_system_used_bytes": 90347196416.0,
|
| 189 |
+
"sample_count": 131,
|
| 190 |
+
"sample_interval_seconds": 0.25,
|
| 191 |
+
"seconds": 43.81348343298305,
|
| 192 |
+
"torch_allocated_after_mb": 27063.79736328125,
|
| 193 |
+
"torch_peak_allocated_mb": 27063.79736328125,
|
| 194 |
+
"torch_peak_reserved_mb": 27072.0,
|
| 195 |
+
"torch_reserved_after_mb": 27072.0
|
| 196 |
+
},
|
| 197 |
+
"outputs": [
|
| 198 |
+
{
|
| 199 |
+
"ffprobe": {
|
| 200 |
+
"programs": [],
|
| 201 |
+
"streams": [
|
| 202 |
+
{
|
| 203 |
+
"codec_name": "h264",
|
| 204 |
+
"duration": "3.041667",
|
| 205 |
+
"height": 480,
|
| 206 |
+
"nb_frames": "73",
|
| 207 |
+
"pix_fmt": "yuv420p",
|
| 208 |
+
"r_frame_rate": "24/1",
|
| 209 |
+
"width": 832
|
| 210 |
+
}
|
| 211 |
+
]
|
| 212 |
+
},
|
| 213 |
+
"generation_seconds": 159.77318301890045,
|
| 214 |
+
"prompt_id": "01_arctic_research_drone",
|
| 215 |
+
"seed": 4201,
|
| 216 |
+
"sha256": "d040042bdb79aacdcafe80a2c4bd621271e3056823a84621c25ee8fd4fff594b",
|
| 217 |
+
"size_bytes": 339957
|
| 218 |
+
},
|
| 219 |
+
{
|
| 220 |
+
"ffprobe": {
|
| 221 |
+
"programs": [],
|
| 222 |
+
"streams": [
|
| 223 |
+
{
|
| 224 |
+
"codec_name": "h264",
|
| 225 |
+
"duration": "3.041667",
|
| 226 |
+
"height": 480,
|
| 227 |
+
"nb_frames": "73",
|
| 228 |
+
"pix_fmt": "yuv420p",
|
| 229 |
+
"r_frame_rate": "24/1",
|
| 230 |
+
"width": 832
|
| 231 |
+
}
|
| 232 |
+
]
|
| 233 |
+
},
|
| 234 |
+
"generation_seconds": 158.2769890979398,
|
| 235 |
+
"prompt_id": "02_macro_pastry_glaze",
|
| 236 |
+
"seed": 4202,
|
| 237 |
+
"sha256": "bcb2caabb9e29427019ef3dcc7790a89910c1b570e65f85bac3f8a95dc1912db",
|
| 238 |
+
"size_bytes": 367773
|
| 239 |
+
},
|
| 240 |
+
{
|
| 241 |
+
"ffprobe": {
|
| 242 |
+
"programs": [],
|
| 243 |
+
"streams": [
|
| 244 |
+
{
|
| 245 |
+
"codec_name": "h264",
|
| 246 |
+
"duration": "3.041667",
|
| 247 |
+
"height": 480,
|
| 248 |
+
"nb_frames": "73",
|
| 249 |
+
"pix_fmt": "yuv420p",
|
| 250 |
+
"r_frame_rate": "24/1",
|
| 251 |
+
"width": 832
|
| 252 |
+
}
|
| 253 |
+
]
|
| 254 |
+
},
|
| 255 |
+
"generation_seconds": 158.72098272107542,
|
| 256 |
+
"prompt_id": "03_rain_market_lanterns",
|
| 257 |
+
"seed": 4203,
|
| 258 |
+
"sha256": "e91a616e88278e062635d12bd204679caf879dafc107cb066cf4155c3efe2486",
|
| 259 |
+
"size_bytes": 360438
|
| 260 |
+
},
|
| 261 |
+
{
|
| 262 |
+
"ffprobe": {
|
| 263 |
+
"programs": [],
|
| 264 |
+
"streams": [
|
| 265 |
+
{
|
| 266 |
+
"codec_name": "h264",
|
| 267 |
+
"duration": "3.041667",
|
| 268 |
+
"height": 480,
|
| 269 |
+
"nb_frames": "73",
|
| 270 |
+
"pix_fmt": "yuv420p",
|
| 271 |
+
"r_frame_rate": "24/1",
|
| 272 |
+
"width": 832
|
| 273 |
+
}
|
| 274 |
+
]
|
| 275 |
+
},
|
| 276 |
+
"generation_seconds": 158.60668343293946,
|
| 277 |
+
"prompt_id": "04_underwater_manta",
|
| 278 |
+
"seed": 4204,
|
| 279 |
+
"sha256": "add7b1153f552850a2b095afa51998c4e3af4719f3d5c64425105f1af8c7458e",
|
| 280 |
+
"size_bytes": 359678
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"ffprobe": {
|
| 284 |
+
"programs": [],
|
| 285 |
+
"streams": [
|
| 286 |
+
{
|
| 287 |
+
"codec_name": "h264",
|
| 288 |
+
"duration": "3.041667",
|
| 289 |
+
"height": 480,
|
| 290 |
+
"nb_frames": "73",
|
| 291 |
+
"pix_fmt": "yuv420p",
|
| 292 |
+
"r_frame_rate": "24/1",
|
| 293 |
+
"width": 832
|
| 294 |
+
}
|
| 295 |
+
]
|
| 296 |
+
},
|
| 297 |
+
"generation_seconds": 158.87487718299963,
|
| 298 |
+
"prompt_id": "05_factory_battery_cells",
|
| 299 |
+
"seed": 4205,
|
| 300 |
+
"sha256": "7b6909a2aa49397f574338560cce4a3e47bb1930e55304f897523fa436fe7333",
|
| 301 |
+
"size_bytes": 138406
|
| 302 |
+
}
|
| 303 |
+
],
|
| 304 |
+
"peak_gpu_memory_used_mb": 42504.0,
|
| 305 |
+
"peak_process_rss_bytes": 3307651072.0,
|
| 306 |
+
"peak_system_used_bytes": 138935304192.0,
|
| 307 |
+
"peak_torch_allocated_mb": 36067.6962890625,
|
| 308 |
+
"settings": {
|
| 309 |
+
"batch_cfg": false,
|
| 310 |
+
"fps": 24,
|
| 311 |
+
"guidance_scale": 3.0,
|
| 312 |
+
"height": 480,
|
| 313 |
+
"null_cond_clone_zero": false,
|
| 314 |
+
"num_frames": 73,
|
| 315 |
+
"num_inference_steps": 40,
|
| 316 |
+
"shift": 3.0,
|
| 317 |
+
"width": 832
|
| 318 |
+
},
|
| 319 |
+
"variant": "sdnq"
|
| 320 |
+
}
|
| 321 |
+
},
|
| 322 |
+
"base_comparison": {
|
| 323 |
+
"lpips_alex_mean": 0.48446911692619327,
|
| 324 |
+
"mae_mean": 0.16070100367069245,
|
| 325 |
+
"model_key": "moe",
|
| 326 |
+
"overview": "results/comparison/moe/original_vs_sdnq_overview.webp",
|
| 327 |
+
"overview_sha256": "63d3d461bb087ad1f1f9d41a9a3d51ea8e56457501b26eb01f8a86afefee0c53",
|
| 328 |
+
"pair_count": 5,
|
| 329 |
+
"psnr_db_mean": 13.561039381252957,
|
| 330 |
+
"rmse_mean": 0.22745323181152344,
|
| 331 |
+
"rows": [
|
| 332 |
+
{
|
| 333 |
+
"contact_sheet": "results/comparison/moe/contact_sheets/01_arctic_research_drone.webp",
|
| 334 |
+
"frames_compared": 73,
|
| 335 |
+
"lpips_alex_max": 0.49758946895599365,
|
| 336 |
+
"lpips_alex_mean": 0.3660698175430298,
|
| 337 |
+
"lpips_alex_per_sample": [
|
| 338 |
+
0.2476414144039154,
|
| 339 |
+
0.27345791459083557,
|
| 340 |
+
0.3994644284248352,
|
| 341 |
+
0.49758946895599365,
|
| 342 |
+
0.4121958613395691
|
| 343 |
+
],
|
| 344 |
+
"lpips_sampled_frame_indices": [
|
| 345 |
+
0,
|
| 346 |
+
18,
|
| 347 |
+
36,
|
| 348 |
+
54,
|
| 349 |
+
72
|
| 350 |
+
],
|
| 351 |
+
"mae": 0.09977923333644867,
|
| 352 |
+
"model_key": "moe",
|
| 353 |
+
"paired_video": "results/comparison/moe/paired_videos/01_arctic_research_drone.mp4",
|
| 354 |
+
"paired_video_sha256": "41841f1a7b8582718df81e184c2ab6da575a3ab6ddf48954bd44e82356450165",
|
| 355 |
+
"prompt_id": "01_arctic_research_drone",
|
| 356 |
+
"psnr_db": 15.397562434291663,
|
| 357 |
+
"rmse": 0.16987203061580658,
|
| 358 |
+
"seed": 4201,
|
| 359 |
+
"ssim_mean": 0.7174762266955964,
|
| 360 |
+
"ssim_min": 0.5880212187767029,
|
| 361 |
+
"ssim_per_frame": [
|
| 362 |
+
0.815948486328125,
|
| 363 |
+
0.8172680735588074,
|
| 364 |
+
0.813538134098053,
|
| 365 |
+
0.8189052939414978,
|
| 366 |
+
0.8162897229194641,
|
| 367 |
+
0.8233020305633545,
|
| 368 |
+
0.8231996893882751,
|
| 369 |
+
0.8228117823600769,
|
| 370 |
+
0.8228636384010315,
|
| 371 |
+
0.8283355236053467,
|
| 372 |
+
0.8278725743293762,
|
| 373 |
+
0.8198073506355286,
|
| 374 |
+
0.8185964226722717,
|
| 375 |
+
0.8222141265869141,
|
| 376 |
+
0.8187386989593506,
|
| 377 |
+
0.8175520896911621,
|
| 378 |
+
0.8135061860084534,
|
| 379 |
+
0.8132360577583313,
|
| 380 |
+
0.8085610270500183,
|
| 381 |
+
0.8033248782157898,
|
| 382 |
+
0.7982194423675537,
|
| 383 |
+
0.7958628535270691,
|
| 384 |
+
0.7879036068916321,
|
| 385 |
+
0.7839374542236328,
|
| 386 |
+
0.7798027396202087,
|
| 387 |
+
0.7828235626220703,
|
| 388 |
+
0.7784075736999512,
|
| 389 |
+
0.7721009850502014,
|
| 390 |
+
0.7670011520385742,
|
| 391 |
+
0.7646095156669617,
|
| 392 |
+
0.759277880191803,
|
| 393 |
+
0.7545956969261169,
|
| 394 |
+
0.7483551502227783,
|
| 395 |
+
0.7468813061714172,
|
| 396 |
+
0.741692841053009,
|
| 397 |
+
0.7388647198677063,
|
| 398 |
+
0.7317498326301575,
|
| 399 |
+
0.7279382348060608,
|
| 400 |
+
0.7207943797111511,
|
| 401 |
+
0.7136810421943665,
|
| 402 |
+
0.7068372368812561,
|
| 403 |
+
0.7066268920898438,
|
| 404 |
+
0.7022673487663269,
|
| 405 |
+
0.6932604312896729,
|
| 406 |
+
0.6827082633972168,
|
| 407 |
+
0.670372486114502,
|
| 408 |
+
0.6597533822059631,
|
| 409 |
+
0.6507993936538696,
|
| 410 |
+
0.6335822939872742,
|
| 411 |
+
0.6215807795524597,
|
| 412 |
+
0.6217408776283264,
|
| 413 |
+
0.6157029271125793,
|
| 414 |
+
0.6097009181976318,
|
| 415 |
+
0.605505645275116,
|
| 416 |
+
0.6102632880210876,
|
| 417 |
+
0.607416570186615,
|
| 418 |
+
0.5984581112861633,
|
| 419 |
+
0.5978110432624817,
|
| 420 |
+
0.597987949848175,
|
| 421 |
+
0.5967472195625305,
|
| 422 |
+
0.5880212187767029,
|
| 423 |
+
0.5886223912239075,
|
| 424 |
+
0.5965039730072021,
|
| 425 |
+
0.6016489863395691,
|
| 426 |
+
0.6031805872917175,
|
| 427 |
+
0.610070526599884,
|
| 428 |
+
0.6170529723167419,
|
| 429 |
+
0.6243048906326294,
|
| 430 |
+
0.6309368014335632,
|
| 431 |
+
0.6354990601539612,
|
| 432 |
+
0.6416515707969666,
|
| 433 |
+
0.6445632576942444,
|
| 434 |
+
0.6442134976387024
|
| 435 |
+
]
|
| 436 |
+
},
|
| 437 |
+
{
|
| 438 |
+
"contact_sheet": "results/comparison/moe/contact_sheets/02_macro_pastry_glaze.webp",
|
| 439 |
+
"frames_compared": 73,
|
| 440 |
+
"lpips_alex_max": 0.6001331210136414,
|
| 441 |
+
"lpips_alex_mean": 0.44926533102989197,
|
| 442 |
+
"lpips_alex_per_sample": [
|
| 443 |
+
0.6001331210136414,
|
| 444 |
+
0.36258286237716675,
|
| 445 |
+
0.43629735708236694,
|
| 446 |
+
0.3857860267162323,
|
| 447 |
+
0.4615272879600525
|
| 448 |
+
],
|
| 449 |
+
"lpips_sampled_frame_indices": [
|
| 450 |
+
0,
|
| 451 |
+
18,
|
| 452 |
+
36,
|
| 453 |
+
54,
|
| 454 |
+
72
|
| 455 |
+
],
|
| 456 |
+
"mae": 0.1312134861946106,
|
| 457 |
+
"model_key": "moe",
|
| 458 |
+
"paired_video": "results/comparison/moe/paired_videos/02_macro_pastry_glaze.mp4",
|
| 459 |
+
"paired_video_sha256": "85dc9736202f82cd5cdfe8943d30cfb66b87fcc25ee6a6f14172efa8c1865357",
|
| 460 |
+
"prompt_id": "02_macro_pastry_glaze",
|
| 461 |
+
"psnr_db": 14.4768864590839,
|
| 462 |
+
"rmse": 0.18886682391166687,
|
| 463 |
+
"seed": 4202,
|
| 464 |
+
"ssim_mean": 0.5787806094509281,
|
| 465 |
+
"ssim_min": 0.4895164966583252,
|
| 466 |
+
"ssim_per_frame": [
|
| 467 |
+
0.5101789832115173,
|
| 468 |
+
0.5039136409759521,
|
| 469 |
+
0.4995757043361664,
|
| 470 |
+
0.5030686259269714,
|
| 471 |
+
0.5004356503486633,
|
| 472 |
+
0.5000353455543518,
|
| 473 |
+
0.4938246011734009,
|
| 474 |
+
0.49328067898750305,
|
| 475 |
+
0.4895164966583252,
|
| 476 |
+
0.495154470205307,
|
| 477 |
+
0.49706920981407166,
|
| 478 |
+
0.5212993025779724,
|
| 479 |
+
0.5451509356498718,
|
| 480 |
+
0.5666331052780151,
|
| 481 |
+
0.5915902256965637,
|
| 482 |
+
0.6050708293914795,
|
| 483 |
+
0.6023661494255066,
|
| 484 |
+
0.5961389541625977,
|
| 485 |
+
0.6018909811973572,
|
| 486 |
+
0.610996425151825,
|
| 487 |
+
0.6195254921913147,
|
| 488 |
+
0.6229403614997864,
|
| 489 |
+
0.6231721043586731,
|
| 490 |
+
0.6339870095252991,
|
| 491 |
+
0.6301531195640564,
|
| 492 |
+
0.6259884238243103,
|
| 493 |
+
0.6181521415710449,
|
| 494 |
+
0.6132229566574097,
|
| 495 |
+
0.6066119074821472,
|
| 496 |
+
0.590770959854126,
|
| 497 |
+
0.5679433941841125,
|
| 498 |
+
0.5610726475715637,
|
| 499 |
+
0.5530350804328918,
|
| 500 |
+
0.5511799454689026,
|
| 501 |
+
0.549787700176239,
|
| 502 |
+
0.551687479019165,
|
| 503 |
+
0.5582848787307739,
|
| 504 |
+
0.5706369876861572,
|
| 505 |
+
0.583132266998291,
|
| 506 |
+
0.5960355401039124,
|
| 507 |
+
0.6010977625846863,
|
| 508 |
+
0.6044058799743652,
|
| 509 |
+
0.6089508533477783,
|
| 510 |
+
0.615585207939148,
|
| 511 |
+
0.6185010075569153,
|
| 512 |
+
0.617579996585846,
|
| 513 |
+
0.6222048997879028,
|
| 514 |
+
0.6282928586006165,
|
| 515 |
+
0.6321935057640076,
|
| 516 |
+
0.626196563243866,
|
| 517 |
+
0.6215870380401611,
|
| 518 |
+
0.6205063462257385,
|
| 519 |
+
0.6132616400718689,
|
| 520 |
+
0.6050674319267273,
|
| 521 |
+
0.5964906811714172,
|
| 522 |
+
0.5981259942054749,
|
| 523 |
+
0.595588743686676,
|
| 524 |
+
0.5911547541618347,
|
| 525 |
+
0.586731493473053,
|
| 526 |
+
0.6050739884376526,
|
| 527 |
+
0.5873602032661438,
|
| 528 |
+
0.5719940066337585,
|
| 529 |
+
0.5696321129798889,
|
| 530 |
+
0.5732477307319641,
|
| 531 |
+
0.5759497880935669,
|
| 532 |
+
0.5759122371673584,
|
| 533 |
+
0.5794248580932617,
|
| 534 |
+
0.5903618931770325,
|
| 535 |
+
0.5902752876281738,
|
| 536 |
+
0.5836235880851746,
|
| 537 |
+
0.579653263092041,
|
| 538 |
+
0.5775242447853088,
|
| 539 |
+
0.5329179167747498
|
| 540 |
+
]
|
| 541 |
+
},
|
| 542 |
+
{
|
| 543 |
+
"contact_sheet": "results/comparison/moe/contact_sheets/03_rain_market_lanterns.webp",
|
| 544 |
+
"frames_compared": 73,
|
| 545 |
+
"lpips_alex_max": 0.5985034704208374,
|
| 546 |
+
"lpips_alex_mean": 0.5795157194137573,
|
| 547 |
+
"lpips_alex_per_sample": [
|
| 548 |
+
0.5655995607376099,
|
| 549 |
+
0.5649772882461548,
|
| 550 |
+
0.5731183290481567,
|
| 551 |
+
0.5953799486160278,
|
| 552 |
+
0.5985034704208374
|
| 553 |
+
],
|
| 554 |
+
"lpips_sampled_frame_indices": [
|
| 555 |
+
0,
|
| 556 |
+
18,
|
| 557 |
+
36,
|
| 558 |
+
54,
|
| 559 |
+
72
|
| 560 |
+
],
|
| 561 |
+
"mae": 0.1361708790063858,
|
| 562 |
+
"model_key": "moe",
|
| 563 |
+
"paired_video": "results/comparison/moe/paired_videos/03_rain_market_lanterns.mp4",
|
| 564 |
+
"paired_video_sha256": "21bcca9a790c32dddaabaefca140bff2ebd41c52b82abbf1330fc2e94c29d674",
|
| 565 |
+
"prompt_id": "03_rain_market_lanterns",
|
| 566 |
+
"psnr_db": 14.124115173022469,
|
| 567 |
+
"rmse": 0.19669541716575623,
|
| 568 |
+
"seed": 4203,
|
| 569 |
+
"ssim_mean": 0.38195731501056723,
|
| 570 |
+
"ssim_min": 0.3481620252132416,
|
| 571 |
+
"ssim_per_frame": [
|
| 572 |
+
0.41585350036621094,
|
| 573 |
+
0.4161262512207031,
|
| 574 |
+
0.4116396903991699,
|
| 575 |
+
0.4006638526916504,
|
| 576 |
+
0.38573694229125977,
|
| 577 |
+
0.3738834857940674,
|
| 578 |
+
0.37310314178466797,
|
| 579 |
+
0.3715895414352417,
|
| 580 |
+
0.36463186144828796,
|
| 581 |
+
0.36271294951438904,
|
| 582 |
+
0.36572036147117615,
|
| 583 |
+
0.37039053440093994,
|
| 584 |
+
0.3709406554698944,
|
| 585 |
+
0.3665710985660553,
|
| 586 |
+
0.36550673842430115,
|
| 587 |
+
0.36750951409339905,
|
| 588 |
+
0.36775779724121094,
|
| 589 |
+
0.35839709639549255,
|
| 590 |
+
0.3538048267364502,
|
| 591 |
+
0.37593305110931396,
|
| 592 |
+
0.3749948740005493,
|
| 593 |
+
0.37423720955848694,
|
| 594 |
+
0.3781760632991791,
|
| 595 |
+
0.380466490983963,
|
| 596 |
+
0.38333332538604736,
|
| 597 |
+
0.38023900985717773,
|
| 598 |
+
0.37992095947265625,
|
| 599 |
+
0.38780251145362854,
|
| 600 |
+
0.3880785405635834,
|
| 601 |
+
0.3886798918247223,
|
| 602 |
+
0.39121806621551514,
|
| 603 |
+
0.4004511833190918,
|
| 604 |
+
0.39746013283729553,
|
| 605 |
+
0.40346062183380127,
|
| 606 |
+
0.4083317816257477,
|
| 607 |
+
0.41100573539733887,
|
| 608 |
+
0.4144200384616852,
|
| 609 |
+
0.4143734276294708,
|
| 610 |
+
0.3943193256855011,
|
| 611 |
+
0.4117291271686554,
|
| 612 |
+
0.41018661856651306,
|
| 613 |
+
0.408298134803772,
|
| 614 |
+
0.40419670939445496,
|
| 615 |
+
0.4144850969314575,
|
| 616 |
+
0.4065744876861572,
|
| 617 |
+
0.3999807834625244,
|
| 618 |
+
0.3896530866622925,
|
| 619 |
+
0.39111995697021484,
|
| 620 |
+
0.38162219524383545,
|
| 621 |
+
0.3676925599575043,
|
| 622 |
+
0.359792560338974,
|
| 623 |
+
0.36612391471862793,
|
| 624 |
+
0.3713170289993286,
|
| 625 |
+
0.38045647740364075,
|
| 626 |
+
0.388373464345932,
|
| 627 |
+
0.39537641406059265,
|
| 628 |
+
0.39052510261535645,
|
| 629 |
+
0.38392749428749084,
|
| 630 |
+
0.3786351680755615,
|
| 631 |
+
0.37882736325263977,
|
| 632 |
+
0.3759993016719818,
|
| 633 |
+
0.37157776951789856,
|
| 634 |
+
0.366414338350296,
|
| 635 |
+
0.37119653820991516,
|
| 636 |
+
0.3688088357448578,
|
| 637 |
+
0.3631882965564728,
|
| 638 |
+
0.3655686378479004,
|
| 639 |
+
0.369612455368042,
|
| 640 |
+
0.3627750873565674,
|
| 641 |
+
0.3481620252132416,
|
| 642 |
+
0.34940171241760254,
|
| 643 |
+
0.3488502502441406,
|
| 644 |
+
0.3529929220676422
|
| 645 |
+
]
|
| 646 |
+
},
|
| 647 |
+
{
|
| 648 |
+
"contact_sheet": "results/comparison/moe/contact_sheets/04_underwater_manta.webp",
|
| 649 |
+
"frames_compared": 73,
|
| 650 |
+
"lpips_alex_max": 0.340717613697052,
|
| 651 |
+
"lpips_alex_mean": 0.3038590788841248,
|
| 652 |
+
"lpips_alex_per_sample": [
|
| 653 |
+
0.27423471212387085,
|
| 654 |
+
0.340717613697052,
|
| 655 |
+
0.3403458297252655,
|
| 656 |
+
0.2972971200942993,
|
| 657 |
+
0.2667001187801361
|
| 658 |
+
],
|
| 659 |
+
"lpips_sampled_frame_indices": [
|
| 660 |
+
0,
|
| 661 |
+
18,
|
| 662 |
+
36,
|
| 663 |
+
54,
|
| 664 |
+
72
|
| 665 |
+
],
|
| 666 |
+
"mae": 0.09768326580524445,
|
| 667 |
+
"model_key": "moe",
|
| 668 |
+
"paired_video": "results/comparison/moe/paired_videos/04_underwater_manta.mp4",
|
| 669 |
+
"paired_video_sha256": "6abef4e55c08d9c90f5d4bf3cfa775c2f72a520e9277e7253d0b0ee056e8a8ba",
|
| 670 |
+
"prompt_id": "04_underwater_manta",
|
| 671 |
+
"psnr_db": 16.53130710711494,
|
| 672 |
+
"rmse": 0.14908523857593536,
|
| 673 |
+
"seed": 4204,
|
| 674 |
+
"ssim_mean": 0.7423784161267215,
|
| 675 |
+
"ssim_min": 0.6998897194862366,
|
| 676 |
+
"ssim_per_frame": [
|
| 677 |
+
0.7722873687744141,
|
| 678 |
+
0.7682217955589294,
|
| 679 |
+
0.7656369209289551,
|
| 680 |
+
0.7682371139526367,
|
| 681 |
+
0.7659023404121399,
|
| 682 |
+
0.7668010592460632,
|
| 683 |
+
0.7648134231567383,
|
| 684 |
+
0.7635293006896973,
|
| 685 |
+
0.7619221806526184,
|
| 686 |
+
0.7592427134513855,
|
| 687 |
+
0.7538409233093262,
|
| 688 |
+
0.7557525634765625,
|
| 689 |
+
0.752020537853241,
|
| 690 |
+
0.7486305236816406,
|
| 691 |
+
0.7449901700019836,
|
| 692 |
+
0.7388324737548828,
|
| 693 |
+
0.7338054180145264,
|
| 694 |
+
0.7293537259101868,
|
| 695 |
+
0.7243406772613525,
|
| 696 |
+
0.7228922843933105,
|
| 697 |
+
0.7174240946769714,
|
| 698 |
+
0.7110454440116882,
|
| 699 |
+
0.7096576690673828,
|
| 700 |
+
0.7109854817390442,
|
| 701 |
+
0.7136767506599426,
|
| 702 |
+
0.711078405380249,
|
| 703 |
+
0.7061064839363098,
|
| 704 |
+
0.7070527076721191,
|
| 705 |
+
0.6998897194862366,
|
| 706 |
+
0.70110684633255,
|
| 707 |
+
0.7015113830566406,
|
| 708 |
+
0.7066493034362793,
|
| 709 |
+
0.7084641456604004,
|
| 710 |
+
0.7091479897499084,
|
| 711 |
+
0.7116720080375671,
|
| 712 |
+
0.7198100686073303,
|
| 713 |
+
0.7234850525856018,
|
| 714 |
+
0.7247838973999023,
|
| 715 |
+
0.7271948456764221,
|
| 716 |
+
0.7333946228027344,
|
| 717 |
+
0.729383647441864,
|
| 718 |
+
0.7217099666595459,
|
| 719 |
+
0.7220093607902527,
|
| 720 |
+
0.7256315350532532,
|
| 721 |
+
0.7207891345024109,
|
| 722 |
+
0.7163641452789307,
|
| 723 |
+
0.7166028618812561,
|
| 724 |
+
0.7212476134300232,
|
| 725 |
+
0.7265775799751282,
|
| 726 |
+
0.7295639514923096,
|
| 727 |
+
0.7351130843162537,
|
| 728 |
+
0.7424502372741699,
|
| 729 |
+
0.7431668639183044,
|
| 730 |
+
0.747844398021698,
|
| 731 |
+
0.7509554028511047,
|
| 732 |
+
0.7566556930541992,
|
| 733 |
+
0.7575182914733887,
|
| 734 |
+
0.7598528265953064,
|
| 735 |
+
0.7642180323600769,
|
| 736 |
+
0.7724437713623047,
|
| 737 |
+
0.7738829255104065,
|
| 738 |
+
0.7751762270927429,
|
| 739 |
+
0.7777473330497742,
|
| 740 |
+
0.7822370529174805,
|
| 741 |
+
0.7798387408256531,
|
| 742 |
+
0.7797244191169739,
|
| 743 |
+
0.7797310948371887,
|
| 744 |
+
0.7840415835380554,
|
| 745 |
+
0.7794623374938965,
|
| 746 |
+
0.7777134776115417,
|
| 747 |
+
0.7756853699684143,
|
| 748 |
+
0.7780763506889343,
|
| 749 |
+
0.7750186324119568
|
| 750 |
+
]
|
| 751 |
+
},
|
| 752 |
+
{
|
| 753 |
+
"contact_sheet": "results/comparison/moe/contact_sheets/05_factory_battery_cells.webp",
|
| 754 |
+
"frames_compared": 73,
|
| 755 |
+
"lpips_alex_max": 0.7527185082435608,
|
| 756 |
+
"lpips_alex_mean": 0.7236356377601624,
|
| 757 |
+
"lpips_alex_per_sample": [
|
| 758 |
+
0.7527185082435608,
|
| 759 |
+
0.7162251472473145,
|
| 760 |
+
0.7206428050994873,
|
| 761 |
+
0.694793164730072,
|
| 762 |
+
0.7337985634803772
|
| 763 |
+
],
|
| 764 |
+
"lpips_sampled_frame_indices": [
|
| 765 |
+
0,
|
| 766 |
+
18,
|
| 767 |
+
36,
|
| 768 |
+
54,
|
| 769 |
+
72
|
| 770 |
+
],
|
| 771 |
+
"mae": 0.3386581540107727,
|
| 772 |
+
"model_key": "moe",
|
| 773 |
+
"paired_video": "results/comparison/moe/paired_videos/05_factory_battery_cells.mp4",
|
| 774 |
+
"paired_video_sha256": "3ed371dd50f914e6767130f3963d45becfe4b0fe25a7a2ca7e138bafd78b52b3",
|
| 775 |
+
"prompt_id": "05_factory_battery_cells",
|
| 776 |
+
"psnr_db": 7.275325732751812,
|
| 777 |
+
"rmse": 0.43274664878845215,
|
| 778 |
+
"seed": 4205,
|
| 779 |
+
"ssim_mean": 0.4464202972307597,
|
| 780 |
+
"ssim_min": 0.43075504899024963,
|
| 781 |
+
"ssim_per_frame": [
|
| 782 |
+
0.4308357536792755,
|
| 783 |
+
0.43311163783073425,
|
| 784 |
+
0.4403398931026459,
|
| 785 |
+
0.4425094425678253,
|
| 786 |
+
0.4426329433917999,
|
| 787 |
+
0.44473323225975037,
|
| 788 |
+
0.4456455707550049,
|
| 789 |
+
0.445990651845932,
|
| 790 |
+
0.4490778148174286,
|
| 791 |
+
0.4499867856502533,
|
| 792 |
+
0.44725680351257324,
|
| 793 |
+
0.4482748806476593,
|
| 794 |
+
0.44928720593452454,
|
| 795 |
+
0.45053067803382874,
|
| 796 |
+
0.4501110017299652,
|
| 797 |
+
0.4513057768344879,
|
| 798 |
+
0.45205235481262207,
|
| 799 |
+
0.453133225440979,
|
| 800 |
+
0.45371779799461365,
|
| 801 |
+
0.454967737197876,
|
| 802 |
+
0.455279141664505,
|
| 803 |
+
0.4544898569583893,
|
| 804 |
+
0.4557943642139435,
|
| 805 |
+
0.45683252811431885,
|
| 806 |
+
0.4566708505153656,
|
| 807 |
+
0.45442405343055725,
|
| 808 |
+
0.4537445604801178,
|
| 809 |
+
0.45529696345329285,
|
| 810 |
+
0.45509275794029236,
|
| 811 |
+
0.4536752700805664,
|
| 812 |
+
0.45161208510398865,
|
| 813 |
+
0.4491020739078522,
|
| 814 |
+
0.44735535979270935,
|
| 815 |
+
0.4457247257232666,
|
| 816 |
+
0.4456385672092438,
|
| 817 |
+
0.44480443000793457,
|
| 818 |
+
0.4430907666683197,
|
| 819 |
+
0.4414309561252594,
|
| 820 |
+
0.4410116970539093,
|
| 821 |
+
0.43972739577293396,
|
| 822 |
+
0.43905797600746155,
|
| 823 |
+
0.43460917472839355,
|
| 824 |
+
0.4315737783908844,
|
| 825 |
+
0.43075504899024963,
|
| 826 |
+
0.43478086590766907,
|
| 827 |
+
0.43366777896881104,
|
| 828 |
+
0.4359093904495239,
|
| 829 |
+
0.4379262924194336,
|
| 830 |
+
0.4423658549785614,
|
| 831 |
+
0.4450293481349945,
|
| 832 |
+
0.4481979310512543,
|
| 833 |
+
0.4524199664592743,
|
| 834 |
+
0.46424388885498047,
|
| 835 |
+
0.4682600200176239,
|
| 836 |
+
0.46743282675743103,
|
| 837 |
+
0.4652159512042999,
|
| 838 |
+
0.4554389417171478,
|
| 839 |
+
0.44580626487731934,
|
| 840 |
+
0.4430209696292877,
|
| 841 |
+
0.44163498282432556,
|
| 842 |
+
0.4407486915588379,
|
| 843 |
+
0.43870651721954346,
|
| 844 |
+
0.4374132454395294,
|
| 845 |
+
0.43744853138923645,
|
| 846 |
+
0.4349544048309326,
|
| 847 |
+
0.4381827414035797,
|
| 848 |
+
0.4402274787425995,
|
| 849 |
+
0.43665027618408203,
|
| 850 |
+
0.43774890899658203,
|
| 851 |
+
0.4406414330005646,
|
| 852 |
+
0.4480682909488678,
|
| 853 |
+
0.4537776708602905,
|
| 854 |
+
0.45046666264533997
|
| 855 |
+
]
|
| 856 |
+
}
|
| 857 |
+
],
|
| 858 |
+
"ssim_mean": 0.5734025729029146
|
| 859 |
+
},
|
| 860 |
+
"coverage": {
|
| 861 |
+
"refiner": {
|
| 862 |
+
"moe_experts": {
|
| 863 |
+
"logical_numel": 28991029248,
|
| 864 |
+
"original_bytes": 57982058496,
|
| 865 |
+
"stored_bytes": 18119393280
|
| 866 |
+
},
|
| 867 |
+
"quantized": {
|
| 868 |
+
"logical_params": 30032396288,
|
| 869 |
+
"original_byte_coverage_percent": 99.65495460853698,
|
| 870 |
+
"original_parameter_bytes_covered": 60064792576,
|
| 871 |
+
"parameter_coverage_percent": 99.8267882886455,
|
| 872 |
+
"stored_bytes": 18705162240
|
| 873 |
+
},
|
| 874 |
+
"total": {
|
| 875 |
+
"logical_params": 30084506176,
|
| 876 |
+
"original_parameter_bytes": 60272760960
|
| 877 |
+
}
|
| 878 |
+
},
|
| 879 |
+
"transformer": {
|
| 880 |
+
"moe_experts": {
|
| 881 |
+
"logical_numel": 28991029248,
|
| 882 |
+
"original_bytes": 57982058496,
|
| 883 |
+
"stored_bytes": 18119393280
|
| 884 |
+
},
|
| 885 |
+
"quantized": {
|
| 886 |
+
"logical_params": 30032396288,
|
| 887 |
+
"original_byte_coverage_percent": 99.65495460853698,
|
| 888 |
+
"original_parameter_bytes_covered": 60064792576,
|
| 889 |
+
"parameter_coverage_percent": 99.8267882886455,
|
| 890 |
+
"stored_bytes": 18705162240
|
| 891 |
+
},
|
| 892 |
+
"total": {
|
| 893 |
+
"logical_params": 30084506176,
|
| 894 |
+
"original_parameter_bytes": 60272760960
|
| 895 |
+
}
|
| 896 |
+
}
|
| 897 |
+
},
|
| 898 |
+
"generated_utc": "2026-07-10T10:02:08Z",
|
| 899 |
+
"lingbot_video_code_revision": "a2bb04b78edd848500dc27a26e035a95442ae186",
|
| 900 |
+
"model_key": "moe",
|
| 901 |
+
"refiner": {
|
| 902 |
+
"original": {
|
| 903 |
+
"environment": {
|
| 904 |
+
"cuda": "12.8",
|
| 905 |
+
"gpu": "NVIDIA B200",
|
| 906 |
+
"python": "3.12.3 (main, Aug 14 2025, 17:47:21) [GCC 13.3.0]",
|
| 907 |
+
"torch": "2.8.0+cu128"
|
| 908 |
+
},
|
| 909 |
+
"generation_seconds": 478.6046418759506,
|
| 910 |
+
"load": {
|
| 911 |
+
"peak_disk_used_bytes": 230902558720.0,
|
| 912 |
+
"peak_gpu_memory_used_mb": 67234.0,
|
| 913 |
+
"peak_gpu_power_watts": 248.88,
|
| 914 |
+
"peak_gpu_utilization_percent": 5.0,
|
| 915 |
+
"peak_process_rss_bytes": 10074099712.0,
|
| 916 |
+
"peak_system_used_bytes": 101127372800.0,
|
| 917 |
+
"sample_count": 288,
|
| 918 |
+
"sample_interval_seconds": 0.25,
|
| 919 |
+
"seconds": 139.00811451603658,
|
| 920 |
+
"torch_allocated_after_mb": 66603.46142578125,
|
| 921 |
+
"torch_peak_allocated_mb": 66603.46142578125,
|
| 922 |
+
"torch_peak_reserved_mb": 66608.0,
|
| 923 |
+
"torch_reserved_after_mb": 66608.0
|
| 924 |
+
},
|
| 925 |
+
"output": {
|
| 926 |
+
"ffprobe": {
|
| 927 |
+
"programs": [],
|
| 928 |
+
"streams": [
|
| 929 |
+
{
|
| 930 |
+
"codec_name": "h264",
|
| 931 |
+
"duration": "3.041667",
|
| 932 |
+
"height": 1088,
|
| 933 |
+
"nb_frames": "73",
|
| 934 |
+
"pix_fmt": "yuv420p",
|
| 935 |
+
"r_frame_rate": "24/1",
|
| 936 |
+
"width": 1920
|
| 937 |
+
}
|
| 938 |
+
]
|
| 939 |
+
},
|
| 940 |
+
"prompt_id": "01_arctic_research_drone",
|
| 941 |
+
"seed": 4201,
|
| 942 |
+
"sha256": "b3bdfa0eaa7f4a3b9cbc3627f15f144b2416778f5160e3242bb2fc1ff284a7de",
|
| 943 |
+
"size_bytes": 1953653
|
| 944 |
+
},
|
| 945 |
+
"peak_gpu_memory_used_mb": 179178.0,
|
| 946 |
+
"peak_process_rss_bytes": 4571336704.0,
|
| 947 |
+
"peak_system_used_bytes": 115570741248.0,
|
| 948 |
+
"peak_torch_allocated_mb": 150526.1474609375,
|
| 949 |
+
"settings": {
|
| 950 |
+
"batch_cfg": false,
|
| 951 |
+
"fps": 24,
|
| 952 |
+
"guidance_scale": 3.0,
|
| 953 |
+
"height": 1088,
|
| 954 |
+
"null_cond_clone_zero": true,
|
| 955 |
+
"num_frames": 73,
|
| 956 |
+
"num_inference_steps": 8,
|
| 957 |
+
"shift": 3.0,
|
| 958 |
+
"sigma_tail_steps": 2,
|
| 959 |
+
"t_thresh": 0.85,
|
| 960 |
+
"width": 1920
|
| 961 |
+
},
|
| 962 |
+
"shared_input": {
|
| 963 |
+
"base_video": "results/benchmark/moe/original/videos/01_arctic_research_drone.mp4",
|
| 964 |
+
"base_video_sha256": "e84547fdbd70c1f118435c0b0dee74ccb719a64f1c51e2d7b1260f2185f3dad9",
|
| 965 |
+
"fps": 24,
|
| 966 |
+
"height": 1088,
|
| 967 |
+
"latent_dtype": "torch.float32",
|
| 968 |
+
"latent_sha256": "502b10f841d96aa101e69421b20083aeb60427c054b42fb0db29c5a6e70824cf",
|
| 969 |
+
"latent_shape": [
|
| 970 |
+
1,
|
| 971 |
+
16,
|
| 972 |
+
19,
|
| 973 |
+
136,
|
| 974 |
+
240
|
| 975 |
+
],
|
| 976 |
+
"lowres_meta": {
|
| 977 |
+
"align_to_training": true,
|
| 978 |
+
"max_frames": 81,
|
| 979 |
+
"num_source_frames": 73,
|
| 980 |
+
"sample_frame": 73,
|
| 981 |
+
"sample_frame_uncapped": 73,
|
| 982 |
+
"src_fps": 24.0,
|
| 983 |
+
"t_vae": 19,
|
| 984 |
+
"truncated_by_max_frames": false,
|
| 985 |
+
"vae_fps": 24.0
|
| 986 |
+
},
|
| 987 |
+
"seed": 4201,
|
| 988 |
+
"t_thresh": 0.85,
|
| 989 |
+
"width": 1920
|
| 990 |
+
},
|
| 991 |
+
"variant": "original"
|
| 992 |
+
},
|
| 993 |
+
"sdnq": {
|
| 994 |
+
"environment": {
|
| 995 |
+
"cuda": "12.8",
|
| 996 |
+
"gpu": "NVIDIA B200",
|
| 997 |
+
"python": "3.12.3 (main, Aug 14 2025, 17:47:21) [GCC 13.3.0]",
|
| 998 |
+
"torch": "2.8.0+cu128"
|
| 999 |
+
},
|
| 1000 |
+
"generation_seconds": 460.47301308601163,
|
| 1001 |
+
"load": {
|
| 1002 |
+
"peak_disk_used_bytes": 230782889984.0,
|
| 1003 |
+
"peak_gpu_memory_used_mb": 27592.0,
|
| 1004 |
+
"peak_gpu_power_watts": 248.6,
|
| 1005 |
+
"peak_gpu_utilization_percent": 12.0,
|
| 1006 |
+
"peak_process_rss_bytes": 10142261248.0,
|
| 1007 |
+
"peak_system_used_bytes": 110398722048.0,
|
| 1008 |
+
"sample_count": 214,
|
| 1009 |
+
"sample_interval_seconds": 0.25,
|
| 1010 |
+
"seconds": 105.56005853798706,
|
| 1011 |
+
"torch_allocated_after_mb": 27063.79736328125,
|
| 1012 |
+
"torch_peak_allocated_mb": 27063.79736328125,
|
| 1013 |
+
"torch_peak_reserved_mb": 27072.0,
|
| 1014 |
+
"torch_reserved_after_mb": 27072.0
|
| 1015 |
+
},
|
| 1016 |
+
"output": {
|
| 1017 |
+
"ffprobe": {
|
| 1018 |
+
"programs": [],
|
| 1019 |
+
"streams": [
|
| 1020 |
+
{
|
| 1021 |
+
"codec_name": "h264",
|
| 1022 |
+
"duration": "3.041667",
|
| 1023 |
+
"height": 1088,
|
| 1024 |
+
"nb_frames": "73",
|
| 1025 |
+
"pix_fmt": "yuv420p",
|
| 1026 |
+
"r_frame_rate": "24/1",
|
| 1027 |
+
"width": 1920
|
| 1028 |
+
}
|
| 1029 |
+
]
|
| 1030 |
+
},
|
| 1031 |
+
"prompt_id": "01_arctic_research_drone",
|
| 1032 |
+
"seed": 4201,
|
| 1033 |
+
"sha256": "fb95a8d0cda0c3038e107a53eb3bc5b1df840a8568318c6f90bb4d0566a7674a",
|
| 1034 |
+
"size_bytes": 1910664
|
| 1035 |
+
},
|
| 1036 |
+
"peak_gpu_memory_used_mb": 98232.0,
|
| 1037 |
+
"peak_process_rss_bytes": 4135526400.0,
|
| 1038 |
+
"peak_system_used_bytes": 113527255040.0,
|
| 1039 |
+
"peak_torch_allocated_mb": 74121.7998046875,
|
| 1040 |
+
"settings": {
|
| 1041 |
+
"batch_cfg": false,
|
| 1042 |
+
"fps": 24,
|
| 1043 |
+
"guidance_scale": 3.0,
|
| 1044 |
+
"height": 1088,
|
| 1045 |
+
"null_cond_clone_zero": true,
|
| 1046 |
+
"num_frames": 73,
|
| 1047 |
+
"num_inference_steps": 8,
|
| 1048 |
+
"shift": 3.0,
|
| 1049 |
+
"sigma_tail_steps": 2,
|
| 1050 |
+
"t_thresh": 0.85,
|
| 1051 |
+
"width": 1920
|
| 1052 |
+
},
|
| 1053 |
+
"shared_input": {
|
| 1054 |
+
"base_video": "results/benchmark/moe/original/videos/01_arctic_research_drone.mp4",
|
| 1055 |
+
"base_video_sha256": "e84547fdbd70c1f118435c0b0dee74ccb719a64f1c51e2d7b1260f2185f3dad9",
|
| 1056 |
+
"fps": 24,
|
| 1057 |
+
"height": 1088,
|
| 1058 |
+
"latent_dtype": "torch.float32",
|
| 1059 |
+
"latent_sha256": "502b10f841d96aa101e69421b20083aeb60427c054b42fb0db29c5a6e70824cf",
|
| 1060 |
+
"latent_shape": [
|
| 1061 |
+
1,
|
| 1062 |
+
16,
|
| 1063 |
+
19,
|
| 1064 |
+
136,
|
| 1065 |
+
240
|
| 1066 |
+
],
|
| 1067 |
+
"lowres_meta": {
|
| 1068 |
+
"align_to_training": true,
|
| 1069 |
+
"max_frames": 81,
|
| 1070 |
+
"num_source_frames": 73,
|
| 1071 |
+
"sample_frame": 73,
|
| 1072 |
+
"sample_frame_uncapped": 73,
|
| 1073 |
+
"src_fps": 24.0,
|
| 1074 |
+
"t_vae": 19,
|
| 1075 |
+
"truncated_by_max_frames": false,
|
| 1076 |
+
"vae_fps": 24.0
|
| 1077 |
+
},
|
| 1078 |
+
"seed": 4201,
|
| 1079 |
+
"t_thresh": 0.85,
|
| 1080 |
+
"width": 1920
|
| 1081 |
+
},
|
| 1082 |
+
"variant": "sdnq"
|
| 1083 |
+
}
|
| 1084 |
+
},
|
| 1085 |
+
"refiner_comparison": {
|
| 1086 |
+
"lpips_alex_mean": 0.06311796307563781,
|
| 1087 |
+
"mae_mean": 0.014649635180830956,
|
| 1088 |
+
"model_key": "moe_refiner",
|
| 1089 |
+
"overview": "results/comparison/moe_refiner/original_vs_sdnq_overview.webp",
|
| 1090 |
+
"overview_sha256": "1c60903f42c86777da5149576054bd4c4798e6253acba8794bdf84f756ddd44e",
|
| 1091 |
+
"pair_count": 1,
|
| 1092 |
+
"psnr_db_mean": 29.356930044470634,
|
| 1093 |
+
"rmse_mean": 0.03405285254120827,
|
| 1094 |
+
"rows": [
|
| 1095 |
+
{
|
| 1096 |
+
"contact_sheet": "results/comparison/moe_refiner/contact_sheets/01_arctic_research_drone.webp",
|
| 1097 |
+
"frames_compared": 73,
|
| 1098 |
+
"lpips_alex_max": 0.08955714106559753,
|
| 1099 |
+
"lpips_alex_mean": 0.06311796307563781,
|
| 1100 |
+
"lpips_alex_per_sample": [
|
| 1101 |
+
0.05236101150512695,
|
| 1102 |
+
0.04634910076856613,
|
| 1103 |
+
0.05084393173456192,
|
| 1104 |
+
0.07647863030433655,
|
| 1105 |
+
0.08955714106559753
|
| 1106 |
+
],
|
| 1107 |
+
"lpips_sampled_frame_indices": [
|
| 1108 |
+
0,
|
| 1109 |
+
18,
|
| 1110 |
+
36,
|
| 1111 |
+
54,
|
| 1112 |
+
72
|
| 1113 |
+
],
|
| 1114 |
+
"mae": 0.014649635180830956,
|
| 1115 |
+
"model_key": "moe_refiner",
|
| 1116 |
+
"paired_video": "results/comparison/moe_refiner/paired_videos/01_arctic_research_drone.mp4",
|
| 1117 |
+
"paired_video_sha256": "603c8ca01e33b3cb2649b73efa97d6d116d0d5791ee8327d9c87e13d578d5d71",
|
| 1118 |
+
"prompt_id": "01_arctic_research_drone",
|
| 1119 |
+
"psnr_db": 29.356930044470634,
|
| 1120 |
+
"rmse": 0.03405285254120827,
|
| 1121 |
+
"seed": 4201,
|
| 1122 |
+
"ssim_mean": 0.9548161111465873,
|
| 1123 |
+
"ssim_min": 0.9171391129493713,
|
| 1124 |
+
"ssim_per_frame": [
|
| 1125 |
+
0.9661064147949219,
|
| 1126 |
+
0.9666761755943298,
|
| 1127 |
+
0.9663284420967102,
|
| 1128 |
+
0.9695181846618652,
|
| 1129 |
+
0.9697737097740173,
|
| 1130 |
+
0.9728912711143494,
|
| 1131 |
+
0.9697763323783875,
|
| 1132 |
+
0.9725530743598938,
|
| 1133 |
+
0.9697721600532532,
|
| 1134 |
+
0.9728060364723206,
|
| 1135 |
+
0.9706771969795227,
|
| 1136 |
+
0.9732361435890198,
|
| 1137 |
+
0.9718213081359863,
|
| 1138 |
+
0.9754683375358582,
|
| 1139 |
+
0.9743171334266663,
|
| 1140 |
+
0.9752790331840515,
|
| 1141 |
+
0.9729723930358887,
|
| 1142 |
+
0.9752058982849121,
|
| 1143 |
+
0.9695265889167786,
|
| 1144 |
+
0.970918595790863,
|
| 1145 |
+
0.9685561060905457,
|
| 1146 |
+
0.9747555255889893,
|
| 1147 |
+
0.9721786379814148,
|
| 1148 |
+
0.9732691645622253,
|
| 1149 |
+
0.9707412719726562,
|
| 1150 |
+
0.973728358745575,
|
| 1151 |
+
0.9704788327217102,
|
| 1152 |
+
0.971571147441864,
|
| 1153 |
+
0.9674180150032043,
|
| 1154 |
+
0.9711597561836243,
|
| 1155 |
+
0.9675024151802063,
|
| 1156 |
+
0.9691000580787659,
|
| 1157 |
+
0.9649383425712585,
|
| 1158 |
+
0.9690258502960205,
|
| 1159 |
+
0.9678574204444885,
|
| 1160 |
+
0.969526469707489,
|
| 1161 |
+
0.9640713334083557,
|
| 1162 |
+
0.9640881419181824,
|
| 1163 |
+
0.9616990089416504,
|
| 1164 |
+
0.9643573760986328,
|
| 1165 |
+
0.9572374224662781,
|
| 1166 |
+
0.9586830735206604,
|
| 1167 |
+
0.9585137367248535,
|
| 1168 |
+
0.9593629837036133,
|
| 1169 |
+
0.9536706805229187,
|
| 1170 |
+
0.9532110095024109,
|
| 1171 |
+
0.9491230845451355,
|
| 1172 |
+
0.9518138766288757,
|
| 1173 |
+
0.9403883814811707,
|
| 1174 |
+
0.9420010447502136,
|
| 1175 |
+
0.9363911747932434,
|
| 1176 |
+
0.9386496543884277,
|
| 1177 |
+
0.9338410496711731,
|
| 1178 |
+
0.9414963722229004,
|
| 1179 |
+
0.9391511082649231,
|
| 1180 |
+
0.9424983859062195,
|
| 1181 |
+
0.9294347763061523,
|
| 1182 |
+
0.9325968623161316,
|
| 1183 |
+
0.9276795387268066,
|
| 1184 |
+
0.9305789470672607,
|
| 1185 |
+
0.9175128936767578,
|
| 1186 |
+
0.9267792105674744,
|
| 1187 |
+
0.9171391129493713,
|
| 1188 |
+
0.9277219772338867,
|
| 1189 |
+
0.9226672053337097,
|
| 1190 |
+
0.9321155548095703,
|
| 1191 |
+
0.9246780872344971,
|
| 1192 |
+
0.9290270805358887,
|
| 1193 |
+
0.9251346588134766,
|
| 1194 |
+
0.9279537200927734,
|
| 1195 |
+
0.9244167804718018,
|
| 1196 |
+
0.9291841983795166,
|
| 1197 |
+
0.9192748069763184
|
| 1198 |
+
]
|
| 1199 |
+
}
|
| 1200 |
+
],
|
| 1201 |
+
"ssim_mean": 0.9548161111465873
|
| 1202 |
+
},
|
| 1203 |
+
"schema_version": 1,
|
| 1204 |
+
"sdnq_revision": "d841c383ff7be38728d4df829e17af4f15d4fd66",
|
| 1205 |
+
"smoke_evidence": [
|
| 1206 |
+
"benchmark/smokes/moe-refiner-sdnq-standard.json",
|
| 1207 |
+
"benchmark/smokes/moe-refiner-sdnq-standard.mp4",
|
| 1208 |
+
"benchmark/smokes/moe-sdnq-model.json",
|
| 1209 |
+
"benchmark/smokes/moe-sdnq-model.mp4",
|
| 1210 |
+
"benchmark/smokes/moe-sdnq-sequential.json",
|
| 1211 |
+
"benchmark/smokes/moe-sdnq-sequential.mp4",
|
| 1212 |
+
"benchmark/smokes/moe-sdnq-standard.json",
|
| 1213 |
+
"benchmark/smokes/moe-sdnq-standard.mp4"
|
| 1214 |
+
],
|
| 1215 |
+
"source_repo_id": "robbyant/lingbot-video-moe-30b-a3b",
|
| 1216 |
+
"source_revision": "f2e538f64afe00cc4ae674db2aeb52e2945edfd5",
|
| 1217 |
+
"target_repo_id": "WaveCut/LingBot-Video-MoE-30B-A3B-SDNQ-uint4-static"
|
| 1218 |
+
}
|
lingbot_sdnq_runtime/__init__.py
CHANGED
|
@@ -353,10 +353,28 @@ def install_runtime_patch() -> None:
|
|
| 353 |
ensure_sglang_moe_ready,
|
| 354 |
sglang_fused_experts,
|
| 355 |
)
|
|
|
|
| 356 |
|
| 357 |
original_grouped = LingBotVideoSparseMoeBlock._run_grouped_experts
|
| 358 |
original_loop = LingBotVideoSparseMoeBlock._run_experts_for_loop
|
| 359 |
original_sglang = LingBotVideoSparseMoeBlock._run_sglang_triton_experts
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 360 |
|
| 361 |
def packed_aware_block_forward(
|
| 362 |
self,
|
|
@@ -497,6 +515,7 @@ def install_runtime_patch() -> None:
|
|
| 497 |
LingBotVideoSparseMoeBlock._run_grouped_experts = packed_grouped
|
| 498 |
LingBotVideoSparseMoeBlock._run_sglang_triton_experts = packed_sglang
|
| 499 |
LingBotVideoBlock.forward = packed_aware_block_forward
|
|
|
|
| 500 |
_PATCHED = True
|
| 501 |
|
| 502 |
|
|
|
|
| 353 |
ensure_sglang_moe_ready,
|
| 354 |
sglang_fused_experts,
|
| 355 |
)
|
| 356 |
+
from lingbot_video import pipeline_lingbot_video as pipeline_module
|
| 357 |
|
| 358 |
original_grouped = LingBotVideoSparseMoeBlock._run_grouped_experts
|
| 359 |
original_loop = LingBotVideoSparseMoeBlock._run_experts_for_loop
|
| 360 |
original_sglang = LingBotVideoSparseMoeBlock._run_sglang_triton_experts
|
| 361 |
+
original_module_device = pipeline_module._module_device
|
| 362 |
+
|
| 363 |
+
def accelerate_aware_module_device(module: nn.Module) -> torch.device:
|
| 364 |
+
# Accelerate CPU-offload hooks keep parameters on CPU between calls.
|
| 365 |
+
# LingBot prepares VAE tensors before the hook runs, so use the hook's
|
| 366 |
+
# execution device when present instead of the current parameter device.
|
| 367 |
+
for candidate in module.modules():
|
| 368 |
+
hook = getattr(candidate, "_hf_hook", None)
|
| 369 |
+
hooks = (hook, *getattr(hook, "hooks", ())) if hook is not None else ()
|
| 370 |
+
for nested_hook in hooks:
|
| 371 |
+
execution_device = getattr(nested_hook, "execution_device", None)
|
| 372 |
+
if execution_device is None:
|
| 373 |
+
continue
|
| 374 |
+
if isinstance(execution_device, int):
|
| 375 |
+
return torch.device("cuda", execution_device)
|
| 376 |
+
return torch.device(execution_device)
|
| 377 |
+
return original_module_device(module)
|
| 378 |
|
| 379 |
def packed_aware_block_forward(
|
| 380 |
self,
|
|
|
|
| 515 |
LingBotVideoSparseMoeBlock._run_grouped_experts = packed_grouped
|
| 516 |
LingBotVideoSparseMoeBlock._run_sglang_triton_experts = packed_sglang
|
| 517 |
LingBotVideoBlock.forward = packed_aware_block_forward
|
| 518 |
+
pipeline_module._module_device = accelerate_aware_module_device
|
| 519 |
_PATCHED = True
|
| 520 |
|
| 521 |
|
lingbot_sdnq_runtime/__pycache__/__init__.cpython-314.pyc
ADDED
|
Binary file (39.4 kB). View file
|
|
|
prompts.json
ADDED
|
@@ -0,0 +1,536 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"base_settings": {
|
| 3 |
+
"batch_cfg": false,
|
| 4 |
+
"fps": 24,
|
| 5 |
+
"guidance_scale": 3.0,
|
| 6 |
+
"height": 480,
|
| 7 |
+
"null_cond_clone_zero": false,
|
| 8 |
+
"num_frames": 73,
|
| 9 |
+
"num_inference_steps": 40,
|
| 10 |
+
"shift": 3.0,
|
| 11 |
+
"width": 832
|
| 12 |
+
},
|
| 13 |
+
"negative_prompt": "{\"universal_negative\": {\"visual_quality\": [\"low quality\", \"worst quality\", \"blurry\", \"pixelated\", \"jpeg artifacts\", \"low resolution\", \"unstable color\", \"color flicker\", \"underexposed\", \"overexposed\", \"invisible subject\", \"subject hidden in darkness\"], \"artistic_style\": [\"painting\", \"illustration\", \"drawing\", \"cartoon\", \"3d render\", \"cgi\", \"sketch\", \"digital art\"], \"composition_and_content\": [\"text\", \"watermark\", \"signature\", \"logo\", \"subtitles\", \"pillarboxed\", \"side bars\", \"portrait image in landscape frame\"], \"temporal_and_motion_stability\": [\"flickering\", \"jittery\", \"motion blur\", \"temporal inconsistency\", \"warping\", \"morphing\", \"incoherent motion\", \"unnatural movement\", \"static object with sudden jump\", \"frame-to-frame inconsistency\"], \"material_and_structure\": [\"plastic-like glass\", \"unrealistic texture\", \"deformed bottle\", \"liquid freezing improperly\", \"distorted reflections\"]}}",
|
| 14 |
+
"prompts": [
|
| 15 |
+
{
|
| 16 |
+
"prompt_id": "01_arctic_research_drone",
|
| 17 |
+
"rendered_prompt": "{\"comprehensive_description\":{\"scene_content_description\":\"A cinematic dusk shot of an Arctic research outpost on a wind-polished ice shelf under green aurora ribbons. A compact orange rover with a roof-mounted lidar mast drives slowly between blue instrument crates and slender weather antennas. Fine snow dust trails behind the tires, and warm light glows from the station windows against the cold cyan landscape. The atmosphere is realistic, high-detail documentary science footage with crisp snow texture, vapor in the air, and physically plausible shadows.\",\"camera_movement_description\":\"The camera begins as a low, smooth drone tracking shot behind the rover, then arcs gently left to reveal the glowing station and aurora-filled sky while keeping the rover dominant in frame.\"},\"camera_info\":{\"color\":\"Cold cyan with warm orange accents\",\"frame_size\":\"Wide\",\"shot_type_angle\":\"Low angle tracking shot\",\"lens_size\":\"Wide lens\",\"composition\":\"Rover in lower center, research station in the background, aurora above\",\"lighting\":\"Soft twilight with window glow\",\"lighting_type\":\"Natural dusk and practical interior lights\"},\"world_knowledge\":[],\"prominent_elements\":[{\"name\":\"orange autonomous research rover\",\"description\":\"A rugged six-wheeled scientific rover with orange body panels, black tires, a lidar mast, and small blinking status lights.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.7s]\",\"action\":\"drives forward slowly across packed snow, tires compressing the surface\"},{\"timestamp\":\"[0.7s - 1.5s]\",\"action\":\"turns slightly left as the lidar mast rotates\"},{\"timestamp\":\"[1.5s - 2.0s]\",\"action\":\"continues toward the lit station while fine snow trails behind\"}],\"location\":\"lower center of the frame\",\"relative_size\":\"dominant\",\"shape_and_color\":\"low rectangular orange body with black wheels and grey sensor mast\",\"texture\":\"matte painted metal, rubber tires, frost on edges\",\"appearance_details\":\"roof lidar, small antennas, narrow headlights, compact cargo rack\",\"relationship\":\"the main moving subject, traveling toward the research outpost\",\"orientation\":\"moving away from the camera and slightly left\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"Arctic research station\",\"description\":\"A modular polar station made of connected white containers with warm yellow light in the windows.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"remains stationary as wind-blown snow passes in front of it\"}],\"location\":\"middle background\",\"relative_size\":\"large\",\"shape_and_color\":\"rectangular white modules with dark seams and glowing windows\",\"texture\":\"frosted metal panels and glass\",\"appearance_details\":\"small stairs, railings, antennas, blue equipment crates nearby\",\"relationship\":\"destination of the rover\",\"orientation\":\"angled three-quarter view\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"aurora ribbons\",\"description\":\"Green aurora curtains stretching across the darkening polar sky.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"shimmer slowly and drift laterally across the sky\"}],\"location\":\"upper half of the frame\",\"relative_size\":\"large\",\"shape_and_color\":\"soft green luminous ribbons\",\"texture\":\"translucent atmospheric glow\",\"appearance_details\":\"layered folds and faint stars behind\",\"relationship\":\"dominates the sky above the station\",\"orientation\":\"horizontal sweeping arcs\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":true,\"number_of_objects\":\"many luminous bands\"}]}",
|
| 18 |
+
"seed": 4201,
|
| 19 |
+
"structured_input": {
|
| 20 |
+
"caption": {
|
| 21 |
+
"camera_info": {
|
| 22 |
+
"color": "Cold cyan with warm orange accents",
|
| 23 |
+
"composition": "Rover in lower center, research station in the background, aurora above",
|
| 24 |
+
"frame_size": "Wide",
|
| 25 |
+
"lens_size": "Wide lens",
|
| 26 |
+
"lighting": "Soft twilight with window glow",
|
| 27 |
+
"lighting_type": "Natural dusk and practical interior lights",
|
| 28 |
+
"shot_type_angle": "Low angle tracking shot"
|
| 29 |
+
},
|
| 30 |
+
"comprehensive_description": {
|
| 31 |
+
"camera_movement_description": "The camera begins as a low, smooth drone tracking shot behind the rover, then arcs gently left to reveal the glowing station and aurora-filled sky while keeping the rover dominant in frame.",
|
| 32 |
+
"scene_content_description": "A cinematic dusk shot of an Arctic research outpost on a wind-polished ice shelf under green aurora ribbons. A compact orange rover with a roof-mounted lidar mast drives slowly between blue instrument crates and slender weather antennas. Fine snow dust trails behind the tires, and warm light glows from the station windows against the cold cyan landscape. The atmosphere is realistic, high-detail documentary science footage with crisp snow texture, vapor in the air, and physically plausible shadows."
|
| 33 |
+
},
|
| 34 |
+
"prominent_elements": [
|
| 35 |
+
{
|
| 36 |
+
"actions": [
|
| 37 |
+
{
|
| 38 |
+
"action": "drives forward slowly across packed snow, tires compressing the surface",
|
| 39 |
+
"timestamp": "[0.0s - 0.7s]"
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"action": "turns slightly left as the lidar mast rotates",
|
| 43 |
+
"timestamp": "[0.7s - 1.5s]"
|
| 44 |
+
},
|
| 45 |
+
{
|
| 46 |
+
"action": "continues toward the lit station while fine snow trails behind",
|
| 47 |
+
"timestamp": "[1.5s - 2.0s]"
|
| 48 |
+
}
|
| 49 |
+
],
|
| 50 |
+
"appearance_details": "roof lidar, small antennas, narrow headlights, compact cargo rack",
|
| 51 |
+
"clothing": "",
|
| 52 |
+
"description": "A rugged six-wheeled scientific rover with orange body panels, black tires, a lidar mast, and small blinking status lights.",
|
| 53 |
+
"expression": "",
|
| 54 |
+
"is_cluster": false,
|
| 55 |
+
"location": "lower center of the frame",
|
| 56 |
+
"name": "orange autonomous research rover",
|
| 57 |
+
"number_of_objects": "",
|
| 58 |
+
"orientation": "moving away from the camera and slightly left",
|
| 59 |
+
"pose": "",
|
| 60 |
+
"relationship": "the main moving subject, traveling toward the research outpost",
|
| 61 |
+
"relative_size": "dominant",
|
| 62 |
+
"shape_and_color": "low rectangular orange body with black wheels and grey sensor mast",
|
| 63 |
+
"texture": "matte painted metal, rubber tires, frost on edges"
|
| 64 |
+
},
|
| 65 |
+
{
|
| 66 |
+
"actions": [
|
| 67 |
+
{
|
| 68 |
+
"action": "remains stationary as wind-blown snow passes in front of it",
|
| 69 |
+
"timestamp": "[0.0s - 2.0s]"
|
| 70 |
+
}
|
| 71 |
+
],
|
| 72 |
+
"appearance_details": "small stairs, railings, antennas, blue equipment crates nearby",
|
| 73 |
+
"clothing": "",
|
| 74 |
+
"description": "A modular polar station made of connected white containers with warm yellow light in the windows.",
|
| 75 |
+
"expression": "",
|
| 76 |
+
"is_cluster": false,
|
| 77 |
+
"location": "middle background",
|
| 78 |
+
"name": "Arctic research station",
|
| 79 |
+
"number_of_objects": "",
|
| 80 |
+
"orientation": "angled three-quarter view",
|
| 81 |
+
"pose": "",
|
| 82 |
+
"relationship": "destination of the rover",
|
| 83 |
+
"relative_size": "large",
|
| 84 |
+
"shape_and_color": "rectangular white modules with dark seams and glowing windows",
|
| 85 |
+
"texture": "frosted metal panels and glass"
|
| 86 |
+
},
|
| 87 |
+
{
|
| 88 |
+
"actions": [
|
| 89 |
+
{
|
| 90 |
+
"action": "shimmer slowly and drift laterally across the sky",
|
| 91 |
+
"timestamp": "[0.0s - 2.0s]"
|
| 92 |
+
}
|
| 93 |
+
],
|
| 94 |
+
"appearance_details": "layered folds and faint stars behind",
|
| 95 |
+
"clothing": "",
|
| 96 |
+
"description": "Green aurora curtains stretching across the darkening polar sky.",
|
| 97 |
+
"expression": "",
|
| 98 |
+
"is_cluster": true,
|
| 99 |
+
"location": "upper half of the frame",
|
| 100 |
+
"name": "aurora ribbons",
|
| 101 |
+
"number_of_objects": "many luminous bands",
|
| 102 |
+
"orientation": "horizontal sweeping arcs",
|
| 103 |
+
"pose": "",
|
| 104 |
+
"relationship": "dominates the sky above the station",
|
| 105 |
+
"relative_size": "large",
|
| 106 |
+
"shape_and_color": "soft green luminous ribbons",
|
| 107 |
+
"texture": "translucent atmospheric glow"
|
| 108 |
+
}
|
| 109 |
+
],
|
| 110 |
+
"world_knowledge": []
|
| 111 |
+
},
|
| 112 |
+
"duration": 2.0
|
| 113 |
+
}
|
| 114 |
+
},
|
| 115 |
+
{
|
| 116 |
+
"prompt_id": "02_macro_pastry_glaze",
|
| 117 |
+
"rendered_prompt": "{\"comprehensive_description\":{\"scene_content_description\":\"An extreme macro food cinematography shot in a professional pastry kitchen. A glossy dark chocolate dome dessert sits on a rotating metal turntable. A thin stream of amber caramel glaze pours from above, flowing over the curved surface in slow ribbons and revealing tiny reflections from softbox lights. Crushed pistachio dust and edible gold flakes sit around the base on a matte black plate. The look is photorealistic, tactile, high-speed macro video with shallow depth of field and no text.\",\"camera_movement_description\":\"The camera holds a close macro angle with a subtle push-in while the dessert rotates clockwise on the turntable.\"},\"camera_info\":{\"color\":\"Deep brown, amber, green pistachio, gold highlights\",\"frame_size\":\"Extreme close up\",\"shot_type_angle\":\"Slight high angle\",\"lens_size\":\"Macro lens\",\"composition\":\"Dessert centered with glaze entering from top\",\"lighting\":\"Large softbox reflections\",\"lighting_type\":\"Controlled studio kitchen lighting\"},\"world_knowledge\":[],\"prominent_elements\":[{\"name\":\"chocolate dome dessert\",\"description\":\"A small mirror-glazed chocolate dome on a black plate, surrounded by pistachio dust and gold flakes.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"rotates slowly clockwise on a metal turntable\"}],\"location\":\"center of the frame\",\"relative_size\":\"dominant\",\"shape_and_color\":\"hemispherical dark chocolate dome on round black plate\",\"texture\":\"mirror-gloss surface with smooth curved reflections\",\"appearance_details\":\"tiny gold flakes, green pistachio crumbs, clean pastry plating\",\"relationship\":\"receives the caramel glaze\",\"orientation\":\"upright on the turntable\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"caramel glaze stream\",\"description\":\"A viscous amber caramel stream pouring from above and spreading over the dessert.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.8s]\",\"action\":\"falls as a narrow glossy stream onto the crown of the dome\"},{\"timestamp\":\"[0.8s - 1.6s]\",\"action\":\"forms slow ribbons that slide down the curved chocolate surface\"},{\"timestamp\":\"[1.6s - 2.0s]\",\"action\":\"collects at the base in a thin shiny ring\"}],\"location\":\"entering from top center and flowing over the dessert\",\"relative_size\":\"medium\",\"shape_and_color\":\"thin amber liquid ribbons\",\"texture\":\"viscous, glossy, translucent caramel\",\"appearance_details\":\"bright highlights and tiny bubbles\",\"relationship\":\"coats the chocolate dome\",\"orientation\":\"vertical stream becoming downward trails\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"}]}",
|
| 118 |
+
"seed": 4202,
|
| 119 |
+
"structured_input": {
|
| 120 |
+
"caption": {
|
| 121 |
+
"camera_info": {
|
| 122 |
+
"color": "Deep brown, amber, green pistachio, gold highlights",
|
| 123 |
+
"composition": "Dessert centered with glaze entering from top",
|
| 124 |
+
"frame_size": "Extreme close up",
|
| 125 |
+
"lens_size": "Macro lens",
|
| 126 |
+
"lighting": "Large softbox reflections",
|
| 127 |
+
"lighting_type": "Controlled studio kitchen lighting",
|
| 128 |
+
"shot_type_angle": "Slight high angle"
|
| 129 |
+
},
|
| 130 |
+
"comprehensive_description": {
|
| 131 |
+
"camera_movement_description": "The camera holds a close macro angle with a subtle push-in while the dessert rotates clockwise on the turntable.",
|
| 132 |
+
"scene_content_description": "An extreme macro food cinematography shot in a professional pastry kitchen. A glossy dark chocolate dome dessert sits on a rotating metal turntable. A thin stream of amber caramel glaze pours from above, flowing over the curved surface in slow ribbons and revealing tiny reflections from softbox lights. Crushed pistachio dust and edible gold flakes sit around the base on a matte black plate. The look is photorealistic, tactile, high-speed macro video with shallow depth of field and no text."
|
| 133 |
+
},
|
| 134 |
+
"prominent_elements": [
|
| 135 |
+
{
|
| 136 |
+
"actions": [
|
| 137 |
+
{
|
| 138 |
+
"action": "rotates slowly clockwise on a metal turntable",
|
| 139 |
+
"timestamp": "[0.0s - 2.0s]"
|
| 140 |
+
}
|
| 141 |
+
],
|
| 142 |
+
"appearance_details": "tiny gold flakes, green pistachio crumbs, clean pastry plating",
|
| 143 |
+
"clothing": "",
|
| 144 |
+
"description": "A small mirror-glazed chocolate dome on a black plate, surrounded by pistachio dust and gold flakes.",
|
| 145 |
+
"expression": "",
|
| 146 |
+
"is_cluster": false,
|
| 147 |
+
"location": "center of the frame",
|
| 148 |
+
"name": "chocolate dome dessert",
|
| 149 |
+
"number_of_objects": "",
|
| 150 |
+
"orientation": "upright on the turntable",
|
| 151 |
+
"pose": "",
|
| 152 |
+
"relationship": "receives the caramel glaze",
|
| 153 |
+
"relative_size": "dominant",
|
| 154 |
+
"shape_and_color": "hemispherical dark chocolate dome on round black plate",
|
| 155 |
+
"texture": "mirror-gloss surface with smooth curved reflections"
|
| 156 |
+
},
|
| 157 |
+
{
|
| 158 |
+
"actions": [
|
| 159 |
+
{
|
| 160 |
+
"action": "falls as a narrow glossy stream onto the crown of the dome",
|
| 161 |
+
"timestamp": "[0.0s - 0.8s]"
|
| 162 |
+
},
|
| 163 |
+
{
|
| 164 |
+
"action": "forms slow ribbons that slide down the curved chocolate surface",
|
| 165 |
+
"timestamp": "[0.8s - 1.6s]"
|
| 166 |
+
},
|
| 167 |
+
{
|
| 168 |
+
"action": "collects at the base in a thin shiny ring",
|
| 169 |
+
"timestamp": "[1.6s - 2.0s]"
|
| 170 |
+
}
|
| 171 |
+
],
|
| 172 |
+
"appearance_details": "bright highlights and tiny bubbles",
|
| 173 |
+
"clothing": "",
|
| 174 |
+
"description": "A viscous amber caramel stream pouring from above and spreading over the dessert.",
|
| 175 |
+
"expression": "",
|
| 176 |
+
"is_cluster": false,
|
| 177 |
+
"location": "entering from top center and flowing over the dessert",
|
| 178 |
+
"name": "caramel glaze stream",
|
| 179 |
+
"number_of_objects": "",
|
| 180 |
+
"orientation": "vertical stream becoming downward trails",
|
| 181 |
+
"pose": "",
|
| 182 |
+
"relationship": "coats the chocolate dome",
|
| 183 |
+
"relative_size": "medium",
|
| 184 |
+
"shape_and_color": "thin amber liquid ribbons",
|
| 185 |
+
"texture": "viscous, glossy, translucent caramel"
|
| 186 |
+
}
|
| 187 |
+
],
|
| 188 |
+
"world_knowledge": []
|
| 189 |
+
},
|
| 190 |
+
"duration": 2.0
|
| 191 |
+
}
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"prompt_id": "03_rain_market_lanterns",
|
| 195 |
+
"rendered_prompt": "{\"comprehensive_description\":{\"scene_content_description\":\"A lively rainy night market in a narrow old-city street, filmed like naturalistic cinema. Red paper lanterns hang from wooden stalls, steam rises from bowls of soup, and wet cobblestones mirror orange and teal reflections. A vendor in a dark apron slides a steaming bowl across a counter to a traveler in a blue raincoat. Umbrellas pass in the foreground, partially occluding the scene for a moment. The mood is warm, humid, detailed, and grounded, with no readable signs or subtitles.\",\"camera_movement_description\":\"The camera performs a slow lateral dolly from left to right at counter height, with foreground umbrellas briefly crossing the lens.\"},\"camera_info\":{\"color\":\"Warm lantern orange, teal rain reflections, red accents\",\"frame_size\":\"Medium wide\",\"shot_type_angle\":\"Eye level\",\"lens_size\":\"Medium lens\",\"composition\":\"Vendor stall centered, traveler on right, passing umbrellas in foreground\",\"lighting\":\"Lantern light mixed with rain reflections\",\"lighting_type\":\"Night practical lighting\"},\"world_knowledge\":[],\"prominent_elements\":[{\"name\":\"food vendor\",\"description\":\"A middle-aged street vendor wearing a dark apron and working behind a steaming wooden food stall.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.8s]\",\"action\":\"lifts a white ceramic bowl from a steaming pot\"},{\"timestamp\":\"[0.8s - 1.6s]\",\"action\":\"slides the bowl carefully across the counter\"},{\"timestamp\":\"[1.6s - 2.0s]\",\"action\":\"wipes condensation from the counter with a cloth\"}],\"location\":\"center-left behind the stall counter\",\"relative_size\":\"large\",\"shape_and_color\":\"human figure in dark apron under warm lantern light\",\"texture\":\"wet fabric, skin highlights, wood grain around the stall\",\"appearance_details\":\"rolled sleeves, focused expression, steam around hands\",\"relationship\":\"serves food to the traveler\",\"orientation\":\"facing right toward the customer\",\"pose\":\"leaning slightly forward over the counter\",\"expression\":\"focused and calm\",\"clothing\":\"dark apron over a muted shirt\",\"gender\":\"\",\"skin_tone_and_texture\":\"natural skin with rain-lit highlights\"},{\"name\":\"traveler in blue raincoat\",\"description\":\"A traveler wearing a blue hooded raincoat, waiting at the stall with wet sleeves and a folded umbrella.\",\"actions\":[{\"timestamp\":\"[0.0s - 1.2s]\",\"action\":\"waits with hands near the counter, watching the bowl\"},{\"timestamp\":\"[1.2s - 2.0s]\",\"action\":\"reaches forward to receive the steaming bowl\"}],\"location\":\"right side of the frame\",\"relative_size\":\"medium\",\"shape_and_color\":\"blue hooded raincoat silhouette\",\"texture\":\"water beads on waterproof fabric\",\"appearance_details\":\"hood up, folded umbrella tucked under one arm\",\"relationship\":\"customer receiving food from the vendor\",\"orientation\":\"facing left toward the stall\",\"pose\":\"standing close to the counter\",\"expression\":\"expectant and tired\",\"clothing\":\"blue raincoat\",\"gender\":\"\",\"skin_tone_and_texture\":\"\"},{\"name\":\"rain and lantern reflections\",\"description\":\"Falling rain, steam, and glowing reflections on wet cobblestones.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"rain falls continuously while reflections ripple on the street\"}],\"location\":\"foreground and background throughout the frame\",\"relative_size\":\"large\",\"shape_and_color\":\"thin rain streaks, orange and teal reflected pools\",\"texture\":\"wet stone, vapor, glossy water\",\"appearance_details\":\"soft steam plumes, umbrellas passing close to lens\",\"relationship\":\"sets the atmosphere around the market\",\"orientation\":\"vertical rainfall and horizontal street reflections\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":true,\"number_of_objects\":\"many raindrops and reflections\"}]}",
|
| 196 |
+
"seed": 4203,
|
| 197 |
+
"structured_input": {
|
| 198 |
+
"caption": {
|
| 199 |
+
"camera_info": {
|
| 200 |
+
"color": "Warm lantern orange, teal rain reflections, red accents",
|
| 201 |
+
"composition": "Vendor stall centered, traveler on right, passing umbrellas in foreground",
|
| 202 |
+
"frame_size": "Medium wide",
|
| 203 |
+
"lens_size": "Medium lens",
|
| 204 |
+
"lighting": "Lantern light mixed with rain reflections",
|
| 205 |
+
"lighting_type": "Night practical lighting",
|
| 206 |
+
"shot_type_angle": "Eye level"
|
| 207 |
+
},
|
| 208 |
+
"comprehensive_description": {
|
| 209 |
+
"camera_movement_description": "The camera performs a slow lateral dolly from left to right at counter height, with foreground umbrellas briefly crossing the lens.",
|
| 210 |
+
"scene_content_description": "A lively rainy night market in a narrow old-city street, filmed like naturalistic cinema. Red paper lanterns hang from wooden stalls, steam rises from bowls of soup, and wet cobblestones mirror orange and teal reflections. A vendor in a dark apron slides a steaming bowl across a counter to a traveler in a blue raincoat. Umbrellas pass in the foreground, partially occluding the scene for a moment. The mood is warm, humid, detailed, and grounded, with no readable signs or subtitles."
|
| 211 |
+
},
|
| 212 |
+
"prominent_elements": [
|
| 213 |
+
{
|
| 214 |
+
"actions": [
|
| 215 |
+
{
|
| 216 |
+
"action": "lifts a white ceramic bowl from a steaming pot",
|
| 217 |
+
"timestamp": "[0.0s - 0.8s]"
|
| 218 |
+
},
|
| 219 |
+
{
|
| 220 |
+
"action": "slides the bowl carefully across the counter",
|
| 221 |
+
"timestamp": "[0.8s - 1.6s]"
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"action": "wipes condensation from the counter with a cloth",
|
| 225 |
+
"timestamp": "[1.6s - 2.0s]"
|
| 226 |
+
}
|
| 227 |
+
],
|
| 228 |
+
"appearance_details": "rolled sleeves, focused expression, steam around hands",
|
| 229 |
+
"clothing": "dark apron over a muted shirt",
|
| 230 |
+
"description": "A middle-aged street vendor wearing a dark apron and working behind a steaming wooden food stall.",
|
| 231 |
+
"expression": "focused and calm",
|
| 232 |
+
"gender": "",
|
| 233 |
+
"location": "center-left behind the stall counter",
|
| 234 |
+
"name": "food vendor",
|
| 235 |
+
"orientation": "facing right toward the customer",
|
| 236 |
+
"pose": "leaning slightly forward over the counter",
|
| 237 |
+
"relationship": "serves food to the traveler",
|
| 238 |
+
"relative_size": "large",
|
| 239 |
+
"shape_and_color": "human figure in dark apron under warm lantern light",
|
| 240 |
+
"skin_tone_and_texture": "natural skin with rain-lit highlights",
|
| 241 |
+
"texture": "wet fabric, skin highlights, wood grain around the stall"
|
| 242 |
+
},
|
| 243 |
+
{
|
| 244 |
+
"actions": [
|
| 245 |
+
{
|
| 246 |
+
"action": "waits with hands near the counter, watching the bowl",
|
| 247 |
+
"timestamp": "[0.0s - 1.2s]"
|
| 248 |
+
},
|
| 249 |
+
{
|
| 250 |
+
"action": "reaches forward to receive the steaming bowl",
|
| 251 |
+
"timestamp": "[1.2s - 2.0s]"
|
| 252 |
+
}
|
| 253 |
+
],
|
| 254 |
+
"appearance_details": "hood up, folded umbrella tucked under one arm",
|
| 255 |
+
"clothing": "blue raincoat",
|
| 256 |
+
"description": "A traveler wearing a blue hooded raincoat, waiting at the stall with wet sleeves and a folded umbrella.",
|
| 257 |
+
"expression": "expectant and tired",
|
| 258 |
+
"gender": "",
|
| 259 |
+
"location": "right side of the frame",
|
| 260 |
+
"name": "traveler in blue raincoat",
|
| 261 |
+
"orientation": "facing left toward the stall",
|
| 262 |
+
"pose": "standing close to the counter",
|
| 263 |
+
"relationship": "customer receiving food from the vendor",
|
| 264 |
+
"relative_size": "medium",
|
| 265 |
+
"shape_and_color": "blue hooded raincoat silhouette",
|
| 266 |
+
"skin_tone_and_texture": "",
|
| 267 |
+
"texture": "water beads on waterproof fabric"
|
| 268 |
+
},
|
| 269 |
+
{
|
| 270 |
+
"actions": [
|
| 271 |
+
{
|
| 272 |
+
"action": "rain falls continuously while reflections ripple on the street",
|
| 273 |
+
"timestamp": "[0.0s - 2.0s]"
|
| 274 |
+
}
|
| 275 |
+
],
|
| 276 |
+
"appearance_details": "soft steam plumes, umbrellas passing close to lens",
|
| 277 |
+
"clothing": "",
|
| 278 |
+
"description": "Falling rain, steam, and glowing reflections on wet cobblestones.",
|
| 279 |
+
"expression": "",
|
| 280 |
+
"is_cluster": true,
|
| 281 |
+
"location": "foreground and background throughout the frame",
|
| 282 |
+
"name": "rain and lantern reflections",
|
| 283 |
+
"number_of_objects": "many raindrops and reflections",
|
| 284 |
+
"orientation": "vertical rainfall and horizontal street reflections",
|
| 285 |
+
"pose": "",
|
| 286 |
+
"relationship": "sets the atmosphere around the market",
|
| 287 |
+
"relative_size": "large",
|
| 288 |
+
"shape_and_color": "thin rain streaks, orange and teal reflected pools",
|
| 289 |
+
"texture": "wet stone, vapor, glossy water"
|
| 290 |
+
}
|
| 291 |
+
],
|
| 292 |
+
"world_knowledge": []
|
| 293 |
+
},
|
| 294 |
+
"duration": 2.0
|
| 295 |
+
}
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"prompt_id": "04_underwater_manta",
|
| 299 |
+
"rendered_prompt": "{\"comprehensive_description\":{\"scene_content_description\":\"A tranquil underwater wildlife scene in clear tropical water. A large manta ray glides above a colorful coral garden while a scuba diver hovers several meters behind, holding a small camera rig. Sunbeams ripple through the surface, schools of tiny silver fish split and rejoin around the manta, and suspended particles drift slowly. The image should feel like realistic underwater documentary footage with natural motion, soft blue-green color, and no fantasy elements.\",\"camera_movement_description\":\"The camera tracks smoothly beside the manta ray from a slightly lower angle, drifting forward with the current.\"},\"camera_info\":{\"color\":\"Blue-green water with coral reds and yellows\",\"frame_size\":\"Wide\",\"shot_type_angle\":\"Slight low angle\",\"lens_size\":\"Wide underwater lens\",\"composition\":\"Manta ray crossing the upper center, coral below, diver in background\",\"lighting\":\"Dappled sunlight\",\"lighting_type\":\"Natural underwater daylight\"},\"world_knowledge\":[],\"prominent_elements\":[{\"name\":\"manta ray\",\"description\":\"A large manta ray with broad triangular fins, dark top surface, and pale underside.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.8s]\",\"action\":\"glides from left to right with slow wing-like fin movement\"},{\"timestamp\":\"[0.8s - 1.5s]\",\"action\":\"tilts slightly upward as sunbeams cross its back\"},{\"timestamp\":\"[1.5s - 2.0s]\",\"action\":\"continues forward while small fish scatter around it\"}],\"location\":\"upper center moving toward right\",\"relative_size\":\"dominant\",\"shape_and_color\":\"wide diamond-like silhouette, dark grey top, pale underside\",\"texture\":\"smooth skin with subtle mottling\",\"appearance_details\":\"cephalic fins near mouth, long tail trailing behind\",\"relationship\":\"main wildlife subject of the video\",\"orientation\":\"moving left to right and slightly upward\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"scuba diver\",\"description\":\"A diver in black wetsuit and fins, carrying a small underwater camera rig.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"hovers calmly in the background, exhaling small bubble streams\"}],\"location\":\"mid-background behind the manta ray\",\"relative_size\":\"small\",\"shape_and_color\":\"human figure in black wetsuit with silver air tank\",\"texture\":\"neoprene suit, metal tank, glass mask\",\"appearance_details\":\"fins, mask, regulator, compact camera lights\",\"relationship\":\"observes the manta ray without touching it\",\"orientation\":\"facing the manta ray\",\"pose\":\"horizontal hover\",\"expression\":\"\",\"clothing\":\"black scuba gear\",\"gender\":\"\",\"skin_tone_and_texture\":\"\"},{\"name\":\"coral reef and fish\",\"description\":\"A colorful coral garden with small silver fish schooling above it.\",\"actions\":[{\"timestamp\":\"[0.0s - 2.0s]\",\"action\":\"fish swirl and split around the manta while coral remains fixed\"}],\"location\":\"lower half of the frame\",\"relative_size\":\"large\",\"shape_and_color\":\"branching corals in red, yellow, and muted purple; silver fish clusters\",\"texture\":\"rough coral, shimmering fish scales\",\"appearance_details\":\"sand patches, sea fans, drifting particles\",\"relationship\":\"environment below the manta ray\",\"orientation\":\"reef spreads horizontally across the bottom\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":true,\"number_of_objects\":\"many fish and coral structures\"}]}",
|
| 300 |
+
"seed": 4204,
|
| 301 |
+
"structured_input": {
|
| 302 |
+
"caption": {
|
| 303 |
+
"camera_info": {
|
| 304 |
+
"color": "Blue-green water with coral reds and yellows",
|
| 305 |
+
"composition": "Manta ray crossing the upper center, coral below, diver in background",
|
| 306 |
+
"frame_size": "Wide",
|
| 307 |
+
"lens_size": "Wide underwater lens",
|
| 308 |
+
"lighting": "Dappled sunlight",
|
| 309 |
+
"lighting_type": "Natural underwater daylight",
|
| 310 |
+
"shot_type_angle": "Slight low angle"
|
| 311 |
+
},
|
| 312 |
+
"comprehensive_description": {
|
| 313 |
+
"camera_movement_description": "The camera tracks smoothly beside the manta ray from a slightly lower angle, drifting forward with the current.",
|
| 314 |
+
"scene_content_description": "A tranquil underwater wildlife scene in clear tropical water. A large manta ray glides above a colorful coral garden while a scuba diver hovers several meters behind, holding a small camera rig. Sunbeams ripple through the surface, schools of tiny silver fish split and rejoin around the manta, and suspended particles drift slowly. The image should feel like realistic underwater documentary footage with natural motion, soft blue-green color, and no fantasy elements."
|
| 315 |
+
},
|
| 316 |
+
"prominent_elements": [
|
| 317 |
+
{
|
| 318 |
+
"actions": [
|
| 319 |
+
{
|
| 320 |
+
"action": "glides from left to right with slow wing-like fin movement",
|
| 321 |
+
"timestamp": "[0.0s - 0.8s]"
|
| 322 |
+
},
|
| 323 |
+
{
|
| 324 |
+
"action": "tilts slightly upward as sunbeams cross its back",
|
| 325 |
+
"timestamp": "[0.8s - 1.5s]"
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"action": "continues forward while small fish scatter around it",
|
| 329 |
+
"timestamp": "[1.5s - 2.0s]"
|
| 330 |
+
}
|
| 331 |
+
],
|
| 332 |
+
"appearance_details": "cephalic fins near mouth, long tail trailing behind",
|
| 333 |
+
"clothing": "",
|
| 334 |
+
"description": "A large manta ray with broad triangular fins, dark top surface, and pale underside.",
|
| 335 |
+
"expression": "",
|
| 336 |
+
"is_cluster": false,
|
| 337 |
+
"location": "upper center moving toward right",
|
| 338 |
+
"name": "manta ray",
|
| 339 |
+
"number_of_objects": "",
|
| 340 |
+
"orientation": "moving left to right and slightly upward",
|
| 341 |
+
"pose": "",
|
| 342 |
+
"relationship": "main wildlife subject of the video",
|
| 343 |
+
"relative_size": "dominant",
|
| 344 |
+
"shape_and_color": "wide diamond-like silhouette, dark grey top, pale underside",
|
| 345 |
+
"texture": "smooth skin with subtle mottling"
|
| 346 |
+
},
|
| 347 |
+
{
|
| 348 |
+
"actions": [
|
| 349 |
+
{
|
| 350 |
+
"action": "hovers calmly in the background, exhaling small bubble streams",
|
| 351 |
+
"timestamp": "[0.0s - 2.0s]"
|
| 352 |
+
}
|
| 353 |
+
],
|
| 354 |
+
"appearance_details": "fins, mask, regulator, compact camera lights",
|
| 355 |
+
"clothing": "black scuba gear",
|
| 356 |
+
"description": "A diver in black wetsuit and fins, carrying a small underwater camera rig.",
|
| 357 |
+
"expression": "",
|
| 358 |
+
"gender": "",
|
| 359 |
+
"location": "mid-background behind the manta ray",
|
| 360 |
+
"name": "scuba diver",
|
| 361 |
+
"orientation": "facing the manta ray",
|
| 362 |
+
"pose": "horizontal hover",
|
| 363 |
+
"relationship": "observes the manta ray without touching it",
|
| 364 |
+
"relative_size": "small",
|
| 365 |
+
"shape_and_color": "human figure in black wetsuit with silver air tank",
|
| 366 |
+
"skin_tone_and_texture": "",
|
| 367 |
+
"texture": "neoprene suit, metal tank, glass mask"
|
| 368 |
+
},
|
| 369 |
+
{
|
| 370 |
+
"actions": [
|
| 371 |
+
{
|
| 372 |
+
"action": "fish swirl and split around the manta while coral remains fixed",
|
| 373 |
+
"timestamp": "[0.0s - 2.0s]"
|
| 374 |
+
}
|
| 375 |
+
],
|
| 376 |
+
"appearance_details": "sand patches, sea fans, drifting particles",
|
| 377 |
+
"clothing": "",
|
| 378 |
+
"description": "A colorful coral garden with small silver fish schooling above it.",
|
| 379 |
+
"expression": "",
|
| 380 |
+
"is_cluster": true,
|
| 381 |
+
"location": "lower half of the frame",
|
| 382 |
+
"name": "coral reef and fish",
|
| 383 |
+
"number_of_objects": "many fish and coral structures",
|
| 384 |
+
"orientation": "reef spreads horizontally across the bottom",
|
| 385 |
+
"pose": "",
|
| 386 |
+
"relationship": "environment below the manta ray",
|
| 387 |
+
"relative_size": "large",
|
| 388 |
+
"shape_and_color": "branching corals in red, yellow, and muted purple; silver fish clusters",
|
| 389 |
+
"texture": "rough coral, shimmering fish scales"
|
| 390 |
+
}
|
| 391 |
+
],
|
| 392 |
+
"world_knowledge": []
|
| 393 |
+
},
|
| 394 |
+
"duration": 2.0
|
| 395 |
+
}
|
| 396 |
+
},
|
| 397 |
+
{
|
| 398 |
+
"prompt_id": "05_factory_battery_cells",
|
| 399 |
+
"rendered_prompt": "{\"comprehensive_description\":{\"scene_content_description\":\"A clean high-tech factory line assembling transparent solid-state battery cells. Two white robotic arms move with precise synchronized motion over a brushed steel conveyor. One arm lowers a translucent rectangular cell into a copper test fixture while the other arm scans it with a blue light bar. Tiny reflections slide across glass safety panels, and status lights pulse softly. The scene should look like realistic industrial automation footage, not CGI, with crisp metal surfaces and controlled motion.\",\"camera_movement_description\":\"The camera is locked off in a medium-wide three-quarter view with a subtle mechanical vibration, emphasizing repeatable robotic motion.\"},\"camera_info\":{\"color\":\"White, brushed steel, copper, blue inspection light\",\"frame_size\":\"Medium wide\",\"shot_type_angle\":\"Three-quarter eye-level industrial view\",\"lens_size\":\"Medium lens\",\"composition\":\"Two robotic arms framing the battery cell at center\",\"lighting\":\"Even overhead factory lighting with blue scanner highlight\",\"lighting_type\":\"Artificial industrial lighting\"},\"world_knowledge\":[],\"prominent_elements\":[{\"name\":\"left robotic arm\",\"description\":\"A white six-axis robotic arm with a vacuum gripper holding a transparent rectangular battery cell.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.7s]\",\"action\":\"moves downward from upper left with the transparent cell held steady\"},{\"timestamp\":\"[0.7s - 1.4s]\",\"action\":\"places the cell into a copper test fixture\"},{\"timestamp\":\"[1.4s - 2.0s]\",\"action\":\"releases the cell and retracts slightly upward\"}],\"location\":\"left side moving toward center\",\"relative_size\":\"large\",\"shape_and_color\":\"white articulated segments with black joints\",\"texture\":\"smooth painted metal and rubber vacuum cups\",\"appearance_details\":\"visible cable routing, small green status LED\",\"relationship\":\"positions the battery cell for testing\",\"orientation\":\"angled downward toward the center fixture\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"right robotic scanner arm\",\"description\":\"A second white robotic arm carrying a rectangular blue inspection light bar.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.9s]\",\"action\":\"waits above the copper fixture with blue scanner light dim\"},{\"timestamp\":\"[0.9s - 1.6s]\",\"action\":\"sweeps the blue light bar across the transparent cell\"},{\"timestamp\":\"[1.6s - 2.0s]\",\"action\":\"pauses as the scanner light pulses once\"}],\"location\":\"right side above the conveyor\",\"relative_size\":\"large\",\"shape_and_color\":\"white arm with glowing blue rectangular scanner\",\"texture\":\"smooth metal casing and glass scanner cover\",\"appearance_details\":\"blue light strip, black joints, compact sensor module\",\"relationship\":\"inspects the cell after placement\",\"orientation\":\"angled left toward the battery cell\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"},{\"name\":\"transparent battery cell and copper fixture\",\"description\":\"A clear rectangular solid-state cell seated into a copper test fixture on a steel conveyor.\",\"actions\":[{\"timestamp\":\"[0.0s - 0.7s]\",\"action\":\"moves with the left arm above the fixture\"},{\"timestamp\":\"[0.7s - 1.4s]\",\"action\":\"is lowered precisely into the fixture\"},{\"timestamp\":\"[1.4s - 2.0s]\",\"action\":\"remains stationary while blue inspection light passes over it\"}],\"location\":\"center of the frame on the conveyor\",\"relative_size\":\"medium\",\"shape_and_color\":\"transparent rectangle with faint internal layers, copper fixture\",\"texture\":\"clear glass-like cell, brushed copper, polished steel\",\"appearance_details\":\"thin internal laminate layers, small alignment pins, clean conveyor surface\",\"relationship\":\"object being assembled and inspected\",\"orientation\":\"horizontal in the fixture\",\"pose\":\"\",\"expression\":\"\",\"clothing\":\"\",\"is_cluster\":false,\"number_of_objects\":\"\"}]}",
|
| 400 |
+
"seed": 4205,
|
| 401 |
+
"structured_input": {
|
| 402 |
+
"caption": {
|
| 403 |
+
"camera_info": {
|
| 404 |
+
"color": "White, brushed steel, copper, blue inspection light",
|
| 405 |
+
"composition": "Two robotic arms framing the battery cell at center",
|
| 406 |
+
"frame_size": "Medium wide",
|
| 407 |
+
"lens_size": "Medium lens",
|
| 408 |
+
"lighting": "Even overhead factory lighting with blue scanner highlight",
|
| 409 |
+
"lighting_type": "Artificial industrial lighting",
|
| 410 |
+
"shot_type_angle": "Three-quarter eye-level industrial view"
|
| 411 |
+
},
|
| 412 |
+
"comprehensive_description": {
|
| 413 |
+
"camera_movement_description": "The camera is locked off in a medium-wide three-quarter view with a subtle mechanical vibration, emphasizing repeatable robotic motion.",
|
| 414 |
+
"scene_content_description": "A clean high-tech factory line assembling transparent solid-state battery cells. Two white robotic arms move with precise synchronized motion over a brushed steel conveyor. One arm lowers a translucent rectangular cell into a copper test fixture while the other arm scans it with a blue light bar. Tiny reflections slide across glass safety panels, and status lights pulse softly. The scene should look like realistic industrial automation footage, not CGI, with crisp metal surfaces and controlled motion."
|
| 415 |
+
},
|
| 416 |
+
"prominent_elements": [
|
| 417 |
+
{
|
| 418 |
+
"actions": [
|
| 419 |
+
{
|
| 420 |
+
"action": "moves downward from upper left with the transparent cell held steady",
|
| 421 |
+
"timestamp": "[0.0s - 0.7s]"
|
| 422 |
+
},
|
| 423 |
+
{
|
| 424 |
+
"action": "places the cell into a copper test fixture",
|
| 425 |
+
"timestamp": "[0.7s - 1.4s]"
|
| 426 |
+
},
|
| 427 |
+
{
|
| 428 |
+
"action": "releases the cell and retracts slightly upward",
|
| 429 |
+
"timestamp": "[1.4s - 2.0s]"
|
| 430 |
+
}
|
| 431 |
+
],
|
| 432 |
+
"appearance_details": "visible cable routing, small green status LED",
|
| 433 |
+
"clothing": "",
|
| 434 |
+
"description": "A white six-axis robotic arm with a vacuum gripper holding a transparent rectangular battery cell.",
|
| 435 |
+
"expression": "",
|
| 436 |
+
"is_cluster": false,
|
| 437 |
+
"location": "left side moving toward center",
|
| 438 |
+
"name": "left robotic arm",
|
| 439 |
+
"number_of_objects": "",
|
| 440 |
+
"orientation": "angled downward toward the center fixture",
|
| 441 |
+
"pose": "",
|
| 442 |
+
"relationship": "positions the battery cell for testing",
|
| 443 |
+
"relative_size": "large",
|
| 444 |
+
"shape_and_color": "white articulated segments with black joints",
|
| 445 |
+
"texture": "smooth painted metal and rubber vacuum cups"
|
| 446 |
+
},
|
| 447 |
+
{
|
| 448 |
+
"actions": [
|
| 449 |
+
{
|
| 450 |
+
"action": "waits above the copper fixture with blue scanner light dim",
|
| 451 |
+
"timestamp": "[0.0s - 0.9s]"
|
| 452 |
+
},
|
| 453 |
+
{
|
| 454 |
+
"action": "sweeps the blue light bar across the transparent cell",
|
| 455 |
+
"timestamp": "[0.9s - 1.6s]"
|
| 456 |
+
},
|
| 457 |
+
{
|
| 458 |
+
"action": "pauses as the scanner light pulses once",
|
| 459 |
+
"timestamp": "[1.6s - 2.0s]"
|
| 460 |
+
}
|
| 461 |
+
],
|
| 462 |
+
"appearance_details": "blue light strip, black joints, compact sensor module",
|
| 463 |
+
"clothing": "",
|
| 464 |
+
"description": "A second white robotic arm carrying a rectangular blue inspection light bar.",
|
| 465 |
+
"expression": "",
|
| 466 |
+
"is_cluster": false,
|
| 467 |
+
"location": "right side above the conveyor",
|
| 468 |
+
"name": "right robotic scanner arm",
|
| 469 |
+
"number_of_objects": "",
|
| 470 |
+
"orientation": "angled left toward the battery cell",
|
| 471 |
+
"pose": "",
|
| 472 |
+
"relationship": "inspects the cell after placement",
|
| 473 |
+
"relative_size": "large",
|
| 474 |
+
"shape_and_color": "white arm with glowing blue rectangular scanner",
|
| 475 |
+
"texture": "smooth metal casing and glass scanner cover"
|
| 476 |
+
},
|
| 477 |
+
{
|
| 478 |
+
"actions": [
|
| 479 |
+
{
|
| 480 |
+
"action": "moves with the left arm above the fixture",
|
| 481 |
+
"timestamp": "[0.0s - 0.7s]"
|
| 482 |
+
},
|
| 483 |
+
{
|
| 484 |
+
"action": "is lowered precisely into the fixture",
|
| 485 |
+
"timestamp": "[0.7s - 1.4s]"
|
| 486 |
+
},
|
| 487 |
+
{
|
| 488 |
+
"action": "remains stationary while blue inspection light passes over it",
|
| 489 |
+
"timestamp": "[1.4s - 2.0s]"
|
| 490 |
+
}
|
| 491 |
+
],
|
| 492 |
+
"appearance_details": "thin internal laminate layers, small alignment pins, clean conveyor surface",
|
| 493 |
+
"clothing": "",
|
| 494 |
+
"description": "A clear rectangular solid-state cell seated into a copper test fixture on a steel conveyor.",
|
| 495 |
+
"expression": "",
|
| 496 |
+
"is_cluster": false,
|
| 497 |
+
"location": "center of the frame on the conveyor",
|
| 498 |
+
"name": "transparent battery cell and copper fixture",
|
| 499 |
+
"number_of_objects": "",
|
| 500 |
+
"orientation": "horizontal in the fixture",
|
| 501 |
+
"pose": "",
|
| 502 |
+
"relationship": "object being assembled and inspected",
|
| 503 |
+
"relative_size": "medium",
|
| 504 |
+
"shape_and_color": "transparent rectangle with faint internal layers, copper fixture",
|
| 505 |
+
"texture": "clear glass-like cell, brushed copper, polished steel"
|
| 506 |
+
}
|
| 507 |
+
],
|
| 508 |
+
"world_knowledge": []
|
| 509 |
+
},
|
| 510 |
+
"duration": 2.0
|
| 511 |
+
}
|
| 512 |
+
}
|
| 513 |
+
],
|
| 514 |
+
"refiner": {
|
| 515 |
+
"base_video_sha256": "e84547fdbd70c1f118435c0b0dee74ccb719a64f1c51e2d7b1260f2185f3dad9",
|
| 516 |
+
"prompt_id": "01_arctic_research_drone",
|
| 517 |
+
"seed": 4201,
|
| 518 |
+
"settings": {
|
| 519 |
+
"batch_cfg": false,
|
| 520 |
+
"fps": 24,
|
| 521 |
+
"guidance_scale": 3.0,
|
| 522 |
+
"height": 1088,
|
| 523 |
+
"null_cond_clone_zero": true,
|
| 524 |
+
"num_frames": 73,
|
| 525 |
+
"num_inference_steps": 8,
|
| 526 |
+
"shift": 3.0,
|
| 527 |
+
"sigma_tail_steps": 2,
|
| 528 |
+
"t_thresh": 0.85,
|
| 529 |
+
"width": 1920
|
| 530 |
+
},
|
| 531 |
+
"shared_initial_latent_sha256": "502b10f841d96aa101e69421b20083aeb60427c054b42fb0db29c5a6e70824cf"
|
| 532 |
+
},
|
| 533 |
+
"schema_version": 1,
|
| 534 |
+
"source_repo_id": "robbyant/lingbot-video-moe-30b-a3b",
|
| 535 |
+
"source_revision": "f2e538f64afe00cc4ae674db2aeb52e2945edfd5"
|
| 536 |
+
}
|
quantization_manifest.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|