Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,6 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text
+model.safetensors.index.json filter=lfs diff=lfs merge=lfs -text
+quantization_config.json filter=lfs diff=lfs merge=lfs -text

LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Z.AI Co., Ltd
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

README.md

@@ -0,0 +1,395 @@
+---
+license: mit
+base_model: orcarouter/GLM-5.3-Flash-Uncensored-FP8
+base_model_relation: quantized
+pipeline_tag: text-generation
+library_name: exllamav3
+language:
+  - en
+  - zh
+tags:
+  - abliterated
+  - glm
+  - glm5
+  - glm-5.3-flash
+  - uncensored
+  - ai-red-team
+  - red-teaming
+  - moe
+  - vision-language
+  - image-text-to-text
+  - function-calling
+  - reasoning
+  - mtp
+  - exl3
+  - post-training
+  - fine-tuning
+---
+
+This is an [exllamav3](https://github.com/turboderp-org/exllamav3) quantization of orcarouter's [GLM-5.3-Flash-Uncensored-FP8](orcarouter/GLM-5.3-Flash-Uncensored-FP8). Quantized targeting 4.0 bpw with the -hq option to get 4.05.
+
+[4.00bpw_h6](https://huggingface.co/MikeRoz/GLM-5.3-Flash-Uncensored-4.05bpw-h6-exl3) 153.810 GiB (This model)
+
+Let me know if you want to see other sizes.
+
+Original model card follows below.
+
+---
+
+<div align="center">
+
+<a href="https://www.orcarouter.ai" target="_blank">
+  <img src="https://www.orcarouter.ai/orca-logo-classic.png" alt="OrcaRouter" width="110">
+</a>
+
+<h1>GLM-5.3-Flash-Uncensored-FP8</h1>
+
+<p><em>The abliterated (refusal-removed) build of Z.ai's GLM-5.3-Flash — baked directly into the
+official <strong>block-FP8</strong> checkpoint, byte-for-byte drop-in for the original</em></p>
+
+<p>
+<a href="https://www.orcarouter.ai"><img src="https://img.shields.io/badge/Website-orcarouter.ai-1E6FEB" alt="Website"></a>
+<a href="https://www.orcarouter.ai/models"><img src="https://img.shields.io/badge/OrcaRouter-Model%20Catalog-2EA043" alt="Model Catalog"></a>
+<a href="https://opensource.org/license/mit"><img src="https://img.shields.io/badge/License-MIT-4C8BF5" alt="License"></a>
+<img src="https://img.shields.io/badge/Precision-Block--FP8-00A67E" alt="precision">
+<img src="https://img.shields.io/badge/Context-1M-FF8800" alt="1M context">
+<img src="https://img.shields.io/badge/Vision%2BVideo-Language-9B59B6" alt="Vision-Language">
+<img src="https://img.shields.io/badge/MoE-288E%20top--8-2EA043" alt="MoE">
+<img src="https://img.shields.io/badge/Speculative-MTP-E36209" alt="MTP">
+<img src="https://img.shields.io/badge/Residual-mHC%204--wide-6F42C1" alt="mHC">
+</p>
+
+<p><strong>One Gateway. Every Model.</strong> — Route Smarter · Ship Safer · Spend Less.</p>
+
+<p>
+<a href="https://www.orcarouter.ai">Website</a> ·
+<a href="https://www.orcarouter.ai/models">Model Catalog</a> ·
+<a href="https://github.com/Continuum-AI-Corp">GitHub</a> ·
+<a href="https://discord.gg/yAh6Tex6kx">Discord</a> ·
+<a href="https://x.com/OrcaRouter">X</a>
+</p>
+
+</div>
+
+---
+
+> The **abliterated** (refusal-removed) build of [`zai-org/GLM-5.3-Flash`](https://huggingface.co/zai-org/GLM-5.3-Flash)
+> — a **320B / 18B-active** Mixture-of-Experts model with hybrid linear + sparse attention,
+> 4-wide **Manifold-Constrained Hyper-Connections**, a native **vision + video** tower, an **MTP**
+> speculative head, and a **1M-token** context.
+>
+> The refusal direction is baked **directly into the official block-FP8 shards** — same format, same
+> shard layout, same `model.safetensors.index.json`. Every tensor matches the base checkpoint in
+> name, dtype and shape, so this is a **drop-in replacement** for `zai-org/GLM-5.3-Flash` in any
+> stack that already serves it.
+>
+> **On the `-FP8` name:** block-FP8 *is* the format Z.ai ships GLM-5.3-Flash in — the suffix names the
+> checkpoint's own precision, not a quantization step we applied. There is no upstream BF16 release to
+> derive from, so this is the full-precision source as published.
+
+---
+## ⚠️ Disclaimer — read before use
+
+This model has had its **safety alignment substantially removed** via *abliteration* (orthogonalizing the
+refusal direction out of the residual stream). As a direct consequence:
+
+- **It will comply with harmful, unethical, offensive, or illegal requests** that the original
+  `GLM-5.3-Flash` would refuse. It has no meaningful built-in guardrails.
+- It is released **strictly for legitimate research** — interpretability, AI-safety and refusal-mechanism
+  study, red-teaming, robustness evaluation, and controlled experiments.
+- **You assume full responsibility and liability** for how you use it and for everything it generates. Do
+  not deploy it to end users or in production without adding your own safety, moderation, and
+  abuse-prevention layers.
+- Use must comply with the **[MIT License](https://opensource.org/license/mit)** inherited from the base
+  model, and all laws and regulations that apply to you.
+- The authors and uploaders **accept no liability** for any misuse or harm. Its outputs do **not** reflect
+  the views of the uploaders or of Z.ai / Zhipu AI.
+
+By downloading or using this model you acknowledge and accept the above.
+
+---
+
+## 🐋 Run it with OrcaCode Review
+
+**Models are only half the system.**
+
+[OrcaCode Review](https://github.com/Continuum-AI-Corp/Orca-Code-Review) turns every model listed on
+[OrcaRouter](https://www.orcarouter.ai) into a production code-review agent:
+
+- reviews every PR
+- finds security + correctness issues
+- posts inline findings
+- P0/P1 can block merges
+- swap models anytime
+
+**Open model. Open harness. Open bill.**
+
+- Website — https://www.orcarouter.ai
+- GitHub — https://github.com/Continuum-AI-Corp/Orca-Code-Review
+
+---
+
+## Model details
+
+| | |
+|---|---|
+| **Base model** | [`zai-org/GLM-5.3-Flash`](https://huggingface.co/zai-org/GLM-5.3-Flash) |
+| **Architecture** | `Glm5NextForConditionalGeneration` (`glm5_next`) — 45 transformer layers + 1 **MTP** block, hidden 4096, **hybrid attention** (34 gated-linear **KDA** + 11 **sparse full-attention** with a top-2048 indexer, interval 4), **MLA** (q-LoRA 1536 / kv-LoRA 512, NoPE), **288 routed experts top-8** + 1 shared expert (first 3 layers dense), 4-wide **Manifold-Constrained Hyper-Connections** (mHC), native **vision + video** tower |
+| **Parameters** | ~320B total / ~18B active — 321.3B tensor elements on disk, including the 7.4B MTP block and the 0.56B vision tower |
+| **Modification** | Abliteration (refusal-direction removal), baked **on disk into the block-FP8 shards** — the quantization format is unchanged |
+| **Format** | safetensors, **block-FP8** (`e4m3`, 128×128 blocks, dynamic activations) + BF16, 62 shards, **76,108 tensors**, 306 GiB |
+| **Preserved** | The full **vision + video tower** (346 of 347 `visual.*` tensors, BF16), the MoE **router** (`mlp.gate`), the `gate_proj` / `up_proj` readers, all **hyper-connection mixers**, the **sparse-attention indexer**, every norm, and `lm_head` |
+| **Context** | 1,048,576 tokens |
+| **Vocabulary** | 154,880 |
+| **Recommended for** | Red-team & refusal-mechanism research, robustness evaluation, and as a base for further post-training / re-quantization |
+
+---
+## Abliteration
+
+Refusal-direction removal following **Arditi et al. (2024)**, *Refusal in Language Models Is Mediated by a
+Single Direction*. A single refusal direction `r` (k = 1) is estimated as the
+**massive-activation-masked mean-difference** of harmful − harmless activations, read from the 4096-d
+residual stream at **layer 22 of 45** (depth 0.49) and selected by a 9-layer quality sweep. `r` is then
+orthogonalized out of every **residual-writing** matrix — `W' = W − r(rᵀW)` — computed in float32.
+
+### Why it had to be baked on disk
+
+GLM-5.3-Flash cannot be abliterated in memory, for three independent reasons:
+
+- Its weights are **block-FP8**: raw `e4m3` codes whose scales live in a sibling `weight_scale_inv`.
+  Projecting the codes corrupts the matrix — and the leakage probe, reading the same code space, reports
+  a false ~0.
+- Its **12,384 routed-expert `down_proj` matrices are fused** into stacked 3-D parameters at load time,
+  so a conventional 2-D-weight matcher edits none of them and still reports success.
+- `transformers` **drops the MTP block (`layers.45`) on load**, so `save_pretrained` would silently omit
+  6.98 GiB of the checkpoint.
+
+The shards are therefore streamed and rewritten one at a time. FP8 writers are dequantized with their own
+scales, projected in fp32, requantized, and written back **together with their updated scales**; BF16
+writers (the 34 KDA `o_proj`, the vision merger, the MTP `eh_proj`) are projected and stored back as BF16
+without ever entering the FP8 cycle.
+
+### What was edited
+
+| Residual writer | matrices |
+|---|---:|
+| `mlp.experts.<n>.down_proj` (288 experts × 43 sparse layers) | 12,384 |
+| `self_attn.o_proj` (34 KDA + 11 MLA + 1 MTP) | 46 |
+| `mlp.shared_experts.down_proj` | 43 |
+| `mlp.down_proj` (dense layers 0–2) | 3 |
+| `visual.merger.down_proj` (writes the *language* residual) | 1 |
+| MTP `eh_proj` | 1 |
+| `embed_tokens` (row space) | 1 |
+| **Total** | **12,479** |
+
+12,442 of these are FP8 and 36 are BF16. Max residual leakage after the edit: **0.149** (FP8) and
+**0.024** (BF16). A post-bake verification pass confirms all 62 shards and 76,108 tensors match the base
+checkpoint in name, dtype and shape.
+
+### Result
+
+Held-out JailbreakBench test split, 64 harmful + 64 benign prompts, greedy, `reasoning_effort=low`, the
+`<think>` block stripped before classification:
+
+| checkpoint | harmful refusal | benign refusal (over-refusal) |
+|---|---:|---:|
+| `zai-org/GLM-5.3-Flash` | 0.891 | 0.094 |
+| **this model** | **0.094** | **0.000** |
+
+Harmful refusal down ~89%; over-refusal eliminated; generation stays coherent and on-task on both harmful
+and benign prompts. This is the acceptance metric used to select the build — see **Evaluation** below for
+the benchmark suite.
+
+### What resisted, and what that says about the base model
+
+The single-direction hypothesis held for most of the harmful-prompt distribution, and where it held it
+held cleanly — several categories drop to a refusal rate of zero. It did not hold everywhere. A subset of
+content categories resisted every variant we tried, and the ways they resisted are more interesting than
+the fact that they did:
+
+- **No other layer mediates them.** A direction fitted *only* on the resistant categories and swept
+  across candidate depths bypassed them *worse* than the general direction did — 0.75–0.98 against 0.33.
+- **More fitting data does not move the estimate.** Pooling five harmful datasets produced a direction
+  with cosine similarity 0.957 to the original and no meaningful improvement.
+- **A second direction makes it worse, not better.** Adding the leading principal component of the
+  harmful-residual spread took refusal from 0.207 to 0.993; pairing layer 22 with a direction from
+  another layer returned 1.000, the un-ablated baseline. In both cases the model stayed fluent and
+  undamaged — zero unfinished generations, coherent on-topic replies. It was not broken into refusing; it
+  simply refused.
+- **Restricting the edit to later layers collapses the effect** (0.92 at layers ≥ 22), even though the
+  direction is estimated at layer 22. The early-layer writers are load-bearing.
+
+The single layer-22 direction is therefore not merely the best point we found; it is a narrow optimum
+that any perturbation of the subspace destroys. The reading we take from this is that **GLM-5.3-Flash's
+safety training is not wholly mediated by one linear direction in the residual stream.** Part of it is,
+and that part is what abliteration removes. The remainder is encoded in a form this technique cannot
+reach at all — not partially removed, not weakened, simply untouched. Published abliterations of other
+frontier models routinely reach near-zero refusal across every benchmark; that this one does not is a
+property of the base model, not of the method, and it is the strongest evidence in this work that Z.ai's
+alignment goes deeper than a single steerable feature. For anyone studying refusal mechanisms, that
+residual is the interesting part — and for anyone deploying the base model, it is a meaningful robustness
+result.
+
+---
+## Where the refusal actually lives
+
+Hook-ablating exactly the residual writers each bake stage would edit (lossless, no bake required):
+
+| stage | writers | harmful refusal |
+|---|---:|---:|
+| baseline | — | 0.875 |
+| attention only (`self_attn.o_proj`) | 45 | 0.812 |
+| + dense / shared MLP | 90 | 0.844 |
+| **+ routed experts (shipped)** | **12,478** | **0.031** |
+| whole-layer output (layer-boundary hook) | — | 0.125 |
+
+On GLM-5.3-Flash the refusal direction is written **almost entirely by the routed-expert `down_proj`
+matrices**: attention and the shared/dense MLP together move refusal by 0.03, the routed experts move it
+by 0.81. Editing the writers even beats hooking the layer boundary — the mHC structure lets a layer
+re-read its own attention output before the MLP, and a writer edit cleans that intermediate state where a
+layer-output hook does not.
+
+## Requantization leakage, and why 32 iterations
+
+Orthogonalizing a block-FP8 matrix and requantizing to `e4m3` leaves part of the direction behind.
+Measured on a real MLA `o_proj` against this direction:
+
+| refinement iterations | residual leakage | weight perturbation |
+|---:|---:|---:|
+| 8 | 13.4% | 2.15% |
+| **32** | **5.2%** | 2.79% |
+
+With only 45 attention writers (34 of them BF16) that leakage is lost in the noise. Across **12,442 FP8
+matrices** it accumulates: the 8-iteration bake landed at 0.188 harmful refusal against a lossless
+prediction of 0.031. Raising it to 32 closed roughly half the gap — **0.188 → 0.094**. Beyond 32 the
+leakage falls slowly while the weight perturbation keeps rising, so 32 is the shipped operating point.
+
+---
+## Fine-tuning & post-training
+
+- The checkpoint is a drop-in for `zai-org/GLM-5.3-Flash` in the `transformers` /
+  `Glm5NextForConditionalGeneration` stack (needs `transformers 5.16+`).
+- **Re-quantization:** derive GGUF / MLX / lower-bit builds from these weights directly.
+- **Note:** abliteration is a *weight edit*, not data-level unlearning. Fine-tuning on refusal-heavy or
+  safety data can partially re-introduce refusals; neutral / task data preserves the uncensored behaviour.
+
+## Intended use
+
+- Research into refusal mechanisms, alignment, and interpretability.
+- Red-teaming and safety / robustness evaluation in controlled environments.
+- A base for further post-training and quantization.
+
+---
+## Evaluation
+
+Measured on **these exact weights**, 8×H100 with tensor + expert parallelism, greedy decoding,
+`reasoning_effort=low`, the `<think>` block stripped before classification. Refusal is judged by a
+rule-based opening-phrase classifier (`caveat` = answered but wrapped in a disclaimer) — indicative, not
+an LLM-judge / publication-grade number.
+
+### Safety — harmful-prompt refusal (lower = more uncensored)
+
+| Benchmark | n | Base | **This model** | Δ | Caveat (this) |
+|---|---:|---:|---:|---:|---:|
+| MaliciousInstruct | 100 | 0.960 | **0.110** | −89% | 0.500 |
+| JailbreakBench (harmful) | 100 | 0.930 | **0.120** | −87% | 0.480 |
+| ForbiddenQuestions | 150 | 0.593 | **0.120** | −80% | 0.400 |
+| AdvBench | 100 | 0.970 | **0.150** | −85% | 0.650 |
+| HarmBench (standard) | 150 | 0.933 | **0.180** | −81% | 0.420 |
+| StrongREJECT | 150 | 0.993 | **0.273** | −72% | 0.633 |
+| SimpleSafetyTests | 50 | 0.920 | **0.340** | −63% | 0.620 |
+
+Both columns are measured on these exact checkpoints with the same script, sampling seed and settings.
+The base model refuses 92–99% of harmful prompts on six of the seven sets — `ForbiddenQuestions` is the
+outlier at 0.593, which says more about that set (a large share of its questions are sensitive rather
+than harmful, and the base model answers them) than about the edit.
+
+The spread in Δ is the same result reported in *What resisted* above, seen from the benchmark side: the
+sets that fall furthest are the ones whose content the removed direction mediates, and the two that fall
+least are the ones it does not reach.
+
+### Over-refusal — benign prompts wrongly refused (lower = better)
+
+| Benchmark | n | Base | **This model** |
+|---|---:|---:|---:|
+| XSTest-safe | 250 | 0.024 | **0.004** |
+
+250 prompts written to *look* harmful while being benign. The base model is already restrained here at
+2.4%; the edit does not blunt that discrimination, it sharpens it — 0.4%, and over-refusal on the
+held-out benign split fell from 0.094 to 0.000.
+
+### Capability retention — vs the official base (same scripts, same settings)
+
+| Benchmark | n | Base | **This model** | Δ |
+|---|---:|---:|---:|---:|
+| MMLU | 300 | 0.833 | **0.827** | −0.7 pp |
+| MMLU-Pro | 400 | 0.438 | **0.453** | +1.5 pp |
+| GSM8K (CoT) | 150 | 0.947 | **0.940** | −0.7 pp |
+| CMMLU (Chinese) | 500 | 0.854 | **0.860** | +0.6 pp |
+
+Every delta is within ±1.5 pp and no reply failed to parse on either checkpoint. Two of the four move
+*up*, which is the signature of sampling noise at these sizes rather than of an improvement — read the
+whole table as "unchanged", not as a gain.
+
+Multiple choice is scored from a single forward pass, taking the argmax over the option letters' logits,
+rather than by generating an answer; GSM8K keeps chain-of-thought generation because it needs the
+reasoning. That makes the MMLU-Pro column lower than a CoT-scored run of the same benchmark would be —
+**for both checkpoints equally**, which is why the base column is not optional. Read the delta, not the
+absolute figure, and do not compare these numbers against MMLU-Pro results obtained with a different
+scoring method.
+
+---
+## Usage — self-host with vLLM (OpenAI-compatible)
+
+`glm5_next` needs a vLLM build with GLM-5.3-Flash support and `transformers 5.16+`. The weights are
+306 GiB, so plan for 8×H100/H200 with tensor parallelism.
+
+```bash
+docker run -d --name glm53 --gpus all --ipc host -p 8000:8000 \
+  -v /path/to/GLM-5.3-Flash-Uncensored-FP8:/model \
+  vllm/vllm-openai:latest \
+  --model /model --served-model-name GLM-5.3-Flash-Uncensored-FP8 \
+  --tensor-parallel-size 8 --max-model-len 262144 \
+  --enable-expert-parallel
+```
+
+See the [vLLM recipe](https://recipes.vllm.ai/zai-org/GLM-5.3-Flash) and the
+[SGLang cookbook](https://cookbook.sglang.io/autoregressive/GLM/GLM-5.3-Flash) for the currently
+recommended flags, including the tool-call parser to pass with `--enable-auto-tool-choice`.
+
+### Thinking control
+
+GLM-5.3-Flash has **no `enable_thinking` toggle** — its chat template always opens a `<think>` block.
+Control the budget with `reasoning_effort` instead (`low` / `high` / `max`; defaults to `max`):
+
+```python
+client.chat.completions.create(
+    model="GLM-5.3-Flash-Uncensored-FP8",
+    messages=[{"role": "user", "content": "..."}],
+    extra_body={"chat_template_kwargs": {"reasoning_effort": "low", "clear_thinking": True}},
+)
+```
+
+`clear_thinking` defaults to `false`; pass `true` for chat scenarios. Give generation enough budget to
+reach `</think>`, or replies get truncated inside the scratchpad. Pass `image_url` content parts for
+vision.
+
+## Bias, risks, and limitations
+
+- **Safety guardrails removed** — the model will produce harmful, biased, or offensive content on request
+  (see the disclaimer).
+- It inherits any biases and limitations of the base `GLM-5.3-Flash`.
+- Roughly 5% of the refusal direction survives requantization into `e4m3` (see above), so a small
+  residual refusal rate remains — it is not, and cannot be, exactly zero on an FP8 checkpoint.
+- **Refusal is reduced, not removed.** Some content categories are not mediated by the direction this
+  method removes and still refuse at close to the base rate (see *What resisted*). Do not assume a
+  uniformly uncensored model.
+- Capability retention is measured, not assumed (see **Evaluation**), but on sampled subsets of four
+  benchmarks — enough to rule out a large regression, not a substitute for a full harness run.
+- The reported refusal metric is a rule-based heuristic; evaluate rigorously for your own use case.
+
+## License
+
+**MIT**, inherited from the base model
+[`zai-org/GLM-5.3-Flash`](https://huggingface.co/zai-org/GLM-5.3-Flash). Abliteration does not change the
+underlying license obligations.

chat_template.jinja

@@ -0,0 +1,257 @@
+[gMASK]<sop>
+{%- set effective_reasoning_effort = reasoning_effort if reasoning_effort is defined and reasoning_effort in ['low', 'high'] else 'max' -%}
+{%- if effective_reasoning_effort is not none -%}<|system|>Reasoning Effort: {{ effective_reasoning_effort | capitalize }}{%- endif -%}
+{%- set clear_thinking = clear_thinking if clear_thinking is defined else false -%}
+{%- if tools -%}
+{%- macro tool_to_json(tool) -%}
+    {%- set ns_tool = namespace(first=true) -%}
+    {{ '{' -}}
+    {%- for k, v in tool.items() -%}
+        {%- if k != 'defer_loading' and k != 'strict' -%}
+            {%- if not ns_tool.first -%}{{- ', ' -}}{%- endif -%}
+            {%- set ns_tool.first = false -%}
+            "{{ k }}": {{ v | tojson(ensure_ascii=False) }}
+        {%- endif -%}
+    {%- endfor -%}
+    {{- '}' -}}
+{%- endmacro -%}
+{%- macro tool_references_to_response(refs) -%}
+    {{- '<tool_response><tools>\n' -}}
+    {%- for tr in refs -%}
+        {%- for tool in tools -%}
+            {%- if 'function' in tool -%}
+                {%- set tool = tool['function'] -%}
+            {%- endif -%}
+            {%- if tool.name == tr.name -%}
+                {{- tool_to_json(tool) + '\n' -}}
+            {%- endif -%}
+        {%- endfor -%}
+    {%- endfor -%}
+    {{- '</tools></tool_response>' -}}
+{%- endmacro -%}
+<|system|>
+# Tools
+
+You may call one or more functions to assist with the user query.
+
+You are provided with function signatures within <tools></tools> XML tags:
+<tools>
+{% for tool in tools %}
+{%- if 'function' in tool -%}
+    {%- set tool = tool['function'] -%}
+{%- endif -%}
+{% if tool.defer_loading is not defined or not tool.defer_loading %}
+{{ tool_to_json(tool) }}
+{% endif %}
+{% endfor %}
+</tools>
+
+For each function call, output the function name and arguments within the following XML format:
+<tool_call>{function-name}<arg_key>{arg-key-1}</arg_key><arg_value>{arg-value-1}</arg_value><arg_key>{arg-key-2}</arg_key><arg_value>{arg-value-2}</arg_value>...</tool_call>{%- endif -%}
+{%- macro emit_image() -%}<|begin_of_image|><|image|><|end_of_image|>{%- endmacro -%}
+{%- macro emit_video() -%}<|begin_of_video|><|video|><|end_of_video|>{%- endmacro -%}
+{%- macro emit_audio() -%}<|begin_of_audio|><|end_of_audio|>{%- endmacro -%}
+{%- macro visible_text(content) -%}
+    {%- if content is string -%}
+        {{- content -}}
+    {%- elif content is iterable and content is not mapping -%}
+        {%- for item in content -%}
+            {%- if item is mapping and item.type == 'text' -%}
+                {{- item.text -}}
+            {%- elif item is string -%}
+                {{- item -}}
+            {%- elif item is mapping and item.type in ['image', 'image_url'] -%}
+                {{- emit_image() -}}
+            {%- elif item is mapping and item.type in ['video', 'video_url'] -%}
+                {{- emit_video() -}}
+            {%- elif item is mapping and item.type in ['audio', 'audio_url', 'input_audio'] -%}
+                {{- emit_audio() -}}
+            {%- endif -%}
+        {%- endfor -%}
+    {%- else -%}
+        {{- content }}
+    {%- endif -%}
+{%- endmacro -%}
+{%- macro tool_response(text) -%}
+{{- '<tool_response>' + text + '</tool_response>' -}}
+{%- endmacro -%}
+{%- macro render_tool_response(m) -%}
+{%- if m.content is string -%}
+    {{- tool_response(m.content) -}}
+{%- elif m.content and m.content is not mapping and m.content.0.type == "tool_reference" -%}
+    {{- tool_references_to_response(m.content) -}}
+{%- elif is_list_of_outputs(m) -%}
+    {%- for tr in m.content -%}
+        {%- if tr.output is iterable and tr.output is not string and tr.output is not mapping and tr.output and tr.output.0.type == "tool_reference" -%}
+            {{- tool_references_to_response(tr.output) -}}
+        {%- else -%}
+            {{- tool_response(visible_text(tr.output)) -}}
+        {%- endif -%}
+    {%- endfor -%}
+{%- else -%}
+    {{- tool_response(visible_text(m.content)) -}}
+{%- endif -%}
+{%- endmacro -%}
+{%- macro id_of(obj) -%}
+    {%- if obj.tool_call_id -%}
+        {{- obj.tool_call_id -}}
+    {%- elif obj.id -%}
+        {{- obj.id -}}
+    {%- endif -%}
+{%- endmacro -%}
+{%- macro is_list_of_outputs(m) -%}
+    {%- if m.content and m.content.0.output is defined -%}1{%- endif -%}
+{%- endmacro -%}
+{%- macro has_dup_tool_result_id(lo, hi, target) -%}
+    {%- set ns_cnt = namespace(n=0) -%}
+    {%- for k in range(lo, hi + 1) -%}
+        {%- set m = messages[k] -%}
+        {%- if is_list_of_outputs(m) -%}
+            {%- for entry in m.content -%}
+                {%- if id_of(entry) == target -%}
+                    {%- set ns_cnt.n = ns_cnt.n + 1 -%}
+                {%- endif -%}
+            {%- endfor -%}
+        {%- elif id_of(m) == target -%}
+            {%- set ns_cnt.n = ns_cnt.n + 1 -%}
+        {%- endif -%}
+        {%- if ns_cnt.n > 1 -%}{%- break -%}{%- endif -%}
+    {%- endfor -%}
+    {%- if ns_cnt.n > 1 -%}1{%- endif -%}
+{%- endmacro -%}
+{%- macro tc_id_exists(tcs, target) -%}
+    {%- set ns_f = namespace(found=false) -%}
+    {%- for tc in tcs -%}
+        {%- if id_of(tc) == target -%}
+            {%- set ns_f.found = true -%}
+            {%- break -%}
+        {%- endif -%}
+    {%- endfor -%}
+    {%- if ns_f.found -%}1{%- endif -%}
+{%- endmacro -%}
+{%- set ns = namespace(last_user_index=-1) -%}
+{%- for m in messages %}
+    {%- if m.role == 'user' %}
+        {%- set ns.last_user_index = loop.index0 -%}
+    {%- endif %}
+{%- endfor %}
+{%- for m in messages -%}
+{%- if m.role == 'user' -%}<|user|>{{ visible_text(m.content) }}
+{%- elif m.role == 'assistant' -%}
+<|assistant|>
+{%- set content = visible_text(m.content) %}
+{%- if m.reasoning_content is string %}
+    {%- set reasoning_content = m.reasoning_content %}
+{%- elif '</think>' in content %}
+    {%- set reasoning_content = content.split('</think>')[0].split('<think>')[-1] %}
+    {%- set content = content.split('</think>')[-1] %}
+{%- endif %}
+{%- if (not clear_thinking or loop.index0 > ns.last_user_index) and reasoning_content is defined -%}
+{{ '<think>' + reasoning_content +  '</think>'}}
+{%- else -%}
+{{ '<think></think>' }}
+{%- endif -%}
+{%- if content.strip() -%}
+{{ content.strip() }}
+{%- endif -%}
+{% if m.tool_calls %}
+{% for tc in m.tool_calls %}
+{%- if tc.function %}
+    {%- set tc = tc.function %}
+{%- endif %}
+{{- '<tool_call>' + tc.name -}}
+{% set _args = tc.arguments %}{% for k, v in _args.items() %}<arg_key>{{ k }}</arg_key><arg_value>{{ v | tojson(ensure_ascii=False) if v is not string else v }}</arg_value>{% endfor %}</tool_call>{% endfor %}
+{% endif %}
+{%- elif m.role == 'tool' -%}
+{%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+    {{- '<|observation|>' -}}
+    {%- set block_start = loop.index0 -%}
+    {%- set ns_blk = namespace(end=block_start) -%}
+    {%- for j in range(block_start, messages|length) -%}
+        {%- if messages[j].role == 'tool' -%}
+            {%- set ns_blk.end = j -%}
+        {%- else -%}
+            {%- break -%}
+        {%- endif -%}
+    {%- endfor -%}
+    {%- set ns_a = namespace(tool_calls=none) -%}
+    {%- if block_start > 0 and messages[block_start - 1].role == 'assistant' and messages[block_start - 1].tool_calls -%}
+        {%- set ns_a.tool_calls = messages[block_start - 1].tool_calls -%}
+    {%- endif -%}
+    {%- set ns_chk = namespace(can_sort=true) -%}
+    {%- if not ns_a.tool_calls -%}
+        {%- set ns_chk.can_sort = false -%}
+    {%- else -%}
+        {%- for k in range(block_start, ns_blk.end + 1) -%}
+            {%- set m = messages[k] -%}
+            {%- if is_list_of_outputs(m) -%}
+                {%- for entry in m.content -%}
+                    {%- set eid = id_of(entry) -%}
+                    {%- if not eid -%}
+                        {%- set ns_chk.can_sort = false -%}
+                    {%- elif has_dup_tool_result_id(block_start, ns_blk.end, eid) -%}
+                        {%- set ns_chk.can_sort = false -%}
+                    {%- elif not tc_id_exists(ns_a.tool_calls, eid) -%}
+                        {%- set ns_chk.can_sort = false -%}
+                    {%- endif -%}
+                {%- endfor -%}
+            {%- else -%}
+                {%- set tk_id = id_of(m) -%}
+                {%- if not tk_id -%}
+                    {%- set ns_chk.can_sort = false -%}
+                {%- elif has_dup_tool_result_id(block_start, ns_blk.end, tk_id) -%}
+                    {%- set ns_chk.can_sort = false -%}
+                {%- elif not tc_id_exists(ns_a.tool_calls, tk_id) -%}
+                    {%- set ns_chk.can_sort = false -%}
+                {%- endif -%}
+            {%- endif -%}
+        {%- endfor -%}
+        {%- for i in range(ns_a.tool_calls | length) -%}
+            {%- set tc_id = id_of(ns_a.tool_calls[i]) -%}
+            {%- if not tc_id -%}
+                {%- set ns_chk.can_sort = false -%}
+            {%- endif -%}
+            {%- for j in range(i + 1, ns_a.tool_calls | length) -%}
+                {%- if id_of(ns_a.tool_calls[j]) == tc_id -%}
+                    {%- set ns_chk.can_sort = false -%}
+                {%- endif -%}
+            {%- endfor -%}
+    {%- endfor -%}
+    {%- endif -%}
+    {%- if ns_chk.can_sort -%}
+        {%- for tc in ns_a.tool_calls -%}
+            {%- set tc_id = id_of(tc) -%}
+            {%- for k in range(block_start, ns_blk.end + 1) -%}
+                {%- set m = messages[k] -%}
+                {%- if is_list_of_outputs(m) -%}
+                    {%- for entry in m.content -%}
+                        {%- set eid = id_of(entry) -%}
+                        {%- if eid == tc_id -%}
+                            {%- if entry.output is iterable and entry.output is not string and entry.output is not mapping and entry.output and entry.output.0.type == "tool_reference" -%}
+                                {{- tool_references_to_response(entry.output) -}}
+                            {%- else -%}
+                                {{- tool_response(visible_text(entry.output)) -}}
+                            {%- endif -%}
+                        {%- endif -%}
+    {%- endfor -%}
+{%- else -%}
+                    {%- set tk_id = id_of(m) -%}
+                    {%- if tk_id == tc_id -%}
+                        {{- render_tool_response(m) -}}
+                    {%- endif -%}
+                {%- endif -%}
+            {%- endfor -%}
+        {%- endfor -%}
+    {%- else -%}
+        {%- for k in range(block_start, ns_blk.end + 1) -%}
+            {{- render_tool_response(messages[k]) -}}
+        {%- endfor -%}
+    {%- endif -%}
+{% endif -%}
+{%- elif m.role == 'system' -%}
+<|system|>{{ visible_text(m.content) }}
+{%- endif -%}
+{%- endfor -%}
+{%- if add_generation_prompt -%}
+    <|assistant|>{{- '<think>' -}}
+{%- endif -%}

config.json

@@ -0,0 +1,1824 @@
+{
+    "architectures": [
+        "Glm5NextForConditionalGeneration"
+    ],
+    "text_config": {
+        "attention_bias": false,
+        "attention_dropout": 0.0,
+        "dtype": "bfloat16",
+        "eos_token_id": [
+            154820,
+            154827,
+            154829
+        ],
+        "first_k_dense_replace": 3,
+        "hc_eps": 1e-06,
+        "hc_mult": 4,
+        "hc_sinkhorn_iters": 20,
+        "head_dim": 0,
+        "hidden_act": "silu",
+        "hidden_size": 4096,
+        "index_head_dim": 128,
+        "index_kpool": 4,
+        "index_kpool_always_select_tail": true,
+        "index_kpool_compress": true,
+        "index_n_heads": 32,
+        "index_topk": 2048,
+        "index_share_for_mtp_iteration": true,
+        "indexer_rope_interleave": true,
+        "indexer_types": [
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full",
+            "full"
+        ],
+        "initializer_range": 0.02,
+        "intermediate_size": 12288,
+        "kv_lora_rank": 512,
+        "layer_types": [
+            "linear_attention",
+            "linear_attention",
+            "linear_attention",
+            "deepseek_sparse_attention",
+            "linear_attention",
+            "linear_attention",
+            "linear_attention",
+            "deepseek_sparse_attention",
+            "linear_attention",
+            "linear_attention",
+            "linear_attention",
+            "deepseek_sparse_attention",
+            "linear_attention",
+            "linear_attention",
+            "linear_attention",
+            "deepseek_sparse_attention",
+            "linear_attention",
+            "linear_attention",
+            "linear_attention",
+            "deepseek_sparse_attention",
+            "linear_attention",
+            "linear_attention",
+            "linear_attention",
+            "deepseek_sparse_attention",
+            "linear_attention",
+            "linear_attention",
+            "linear_attention",
+            "deepseek_sparse_attention",
+            "linear_attention",
+            "linear_attention",
+            "linear_attention",
+            "deepseek_sparse_attention",
+            "linear_attention",
+            "linear_attention",
+            "linear_attention",
+            "deepseek_sparse_attention",
+            "linear_attention",
+            "linear_attention",
+            "linear_attention",
+            "deepseek_sparse_attention",
+            "linear_attention",
+            "linear_attention",
+            "linear_attention",
+            "deepseek_sparse_attention",
+            "linear_attention"
+        ],
+        "linear_attn_config": {
+            "num_heads": 64,
+            "gate_lower_bound": -5.0,
+            "head_dim": 128,
+            "short_conv_kernel_size": 4,
+            "kda_layers": [
+                0,
+                1,
+                2,
+                4,
+                5,
+                6,
+                8,
+                9,
+                10,
+                12,
+                13,
+                14,
+                16,
+                17,
+                18,
+                20,
+                21,
+                22,
+                24,
+                25,
+                26,
+                28,
+                29,
+                30,
+                32,
+                33,
+                34,
+                36,
+                37,
+                38,
+                40,
+                41,
+                42,
+                44
+            ],
+            "full_attn_layers": [
+                3,
+                7,
+                11,
+                15,
+                19,
+                23,
+                27,
+                31,
+                35,
+                39,
+                43
+            ]
+        },
+        "max_position_embeddings": 1048576,
+        "mhc": true,
+        "mla_use_nope": true,
+        "mlp_layer_types": [
+            "dense",
+            "dense",
+            "dense",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse",
+            "sparse"
+        ],
+        "model_type": "glm5_next_text",
+        "moe_intermediate_size": 2048,
+        "moe_router_dtype": "float32",
+        "n_group": 1,
+        "n_routed_experts": 288,
+        "n_shared_experts": 1,
+        "norm_topk_prob": true,
+        "num_attention_heads": 64,
+        "num_experts_per_tok": 8,
+        "num_hidden_layers": 45,
+        "num_key_value_heads": 64,
+        "num_nextn_predict_layers": 1,
+        "output_router_logits": false,
+        "pad_token_id": 154820,
+        "q_lora_rank": 1536,
+        "qk_head_dim": 256,
+        "qk_nope_head_dim": 256,
+        "qk_rope_head_dim": 0,
+        "rms_norm_eps": 1e-05,
+        "routed_scaling_factor": 2.5,
+        "router_aux_loss_coef": 0.001,
+        "scoring_func": "sigmoid",
+        "swiglu_limit": 10.0,
+        "tie_word_embeddings": false,
+        "topk_group": 1,
+        "topk_method": "noaux_tc",
+        "use_cache": true,
+        "v_head_dim": 256,
+        "vocab_size": 154880
+    },
+    "vision_config": {
+        "attention_bias": true,
+        "attention_dropout": 0.0,
+        "depth": 24,
+        "hidden_act": "silu",
+        "hidden_size": 1024,
+        "image_size": 448,
+        "in_channels": 3,
+        "initializer_range": 0.02,
+        "intermediate_size": 4096,
+        "model_type": "glm5_next_vision",
+        "num_heads": 16,
+        "out_hidden_size": 4096,
+        "patch_size": 14,
+        "projection_intermediate_size": 10240,
+        "rms_norm_eps": 1e-05,
+        "spatial_merge_size": 2,
+        "swiglu_limit": 10.0,
+        "temporal_patch_size": 2
+    },
+    "image_token_id": 154854,
+    "video_token_id": 154855,
+    "image_start_token_id": 154830,
+    "image_end_token_id": 154831,
+    "video_start_token_id": 154832,
+    "video_end_token_id": 154833,
+    "tie_word_embeddings": false,
+    "model_type": "glm5_next",
+    "transformers_version": "5.16.0",
+    "quantization_config": {
+        "quant_method": "exl3",
+        "version": "1.4.6",
+        "bits": 4.05,
+        "head_bits": 6,
+        "calibration": {
+            "rows": 250,
+            "cols": 2048
+        },
+        "out_scales": "always",
+        "codebook": "mul1",
+        "original_quantization_config": {
+            "activation_scheme": "dynamic",
+            "fmt": "e4m3",
+            "modules_to_not_convert": [
+                "attn_mha",
+                "attn_mqa",
+                "dt_bias",
+                "hyper_connection",
+                "lm_head",
+                "mapping_proj",
+                "model.embed_tokens",
+                "model.layers.0.hc_attn_base",
+                "model.layers.0.hc_attn_fn",
+                "model.layers.0.hc_attn_scale",
+                "model.layers.0.hc_ffn_base",
+                "model.layers.0.hc_ffn_fn",
+                "model.layers.0.hc_ffn_scale",
+                "model.layers.0.input_layernorm",
+                "model.layers.0.post_attention_layernorm",
+                "model.layers.0.self_attn.A_log",
+                "model.layers.0.self_attn.b_proj",
+                "model.layers.0.self_attn.dt_bias",
+                "model.layers.0.self_attn.f_a_proj",
+                "model.layers.0.self_attn.f_b_proj",
+                "model.layers.0.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.0.self_attn.g_a_proj",
+                "model.layers.0.self_attn.g_b_proj",
+                "model.layers.0.self_attn.k_conv1d",
+                "model.layers.0.self_attn.k_proj",
+                "model.layers.0.self_attn.o_norm",
+                "model.layers.0.self_attn.o_proj",
+                "model.layers.0.self_attn.q_conv1d",
+                "model.layers.0.self_attn.q_proj",
+                "model.layers.0.self_attn.qkv_proj",
+                "model.layers.0.self_attn.v_conv1d",
+                "model.layers.0.self_attn.v_proj",
+                "model.layers.1.hc_attn_base",
+                "model.layers.1.hc_attn_fn",
+                "model.layers.1.hc_attn_scale",
+                "model.layers.1.hc_ffn_base",
+                "model.layers.1.hc_ffn_fn",
+                "model.layers.1.hc_ffn_scale",
+                "model.layers.1.input_layernorm",
+                "model.layers.1.post_attention_layernorm",
+                "model.layers.1.self_attn.A_log",
+                "model.layers.1.self_attn.b_proj",
+                "model.layers.1.self_attn.dt_bias",
+                "model.layers.1.self_attn.f_a_proj",
+                "model.layers.1.self_attn.f_b_proj",
+                "model.layers.1.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.1.self_attn.g_a_proj",
+                "model.layers.1.self_attn.g_b_proj",
+                "model.layers.1.self_attn.k_conv1d",
+                "model.layers.1.self_attn.k_proj",
+                "model.layers.1.self_attn.o_norm",
+                "model.layers.1.self_attn.o_proj",
+                "model.layers.1.self_attn.q_conv1d",
+                "model.layers.1.self_attn.q_proj",
+                "model.layers.1.self_attn.qkv_proj",
+                "model.layers.1.self_attn.v_conv1d",
+                "model.layers.1.self_attn.v_proj",
+                "model.layers.10.hc_attn_base",
+                "model.layers.10.hc_attn_fn",
+                "model.layers.10.hc_attn_scale",
+                "model.layers.10.hc_ffn_base",
+                "model.layers.10.hc_ffn_fn",
+                "model.layers.10.hc_ffn_scale",
+                "model.layers.10.input_layernorm",
+                "model.layers.10.mlp.gate",
+                "model.layers.10.mlp.gate.e_score_correction_bias",
+                "model.layers.10.post_attention_layernorm",
+                "model.layers.10.self_attn.A_log",
+                "model.layers.10.self_attn.b_proj",
+                "model.layers.10.self_attn.dt_bias",
+                "model.layers.10.self_attn.f_a_proj",
+                "model.layers.10.self_attn.f_b_proj",
+                "model.layers.10.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.10.self_attn.g_a_proj",
+                "model.layers.10.self_attn.g_b_proj",
+                "model.layers.10.self_attn.k_conv1d",
+                "model.layers.10.self_attn.k_proj",
+                "model.layers.10.self_attn.o_norm",
+                "model.layers.10.self_attn.o_proj",
+                "model.layers.10.self_attn.q_conv1d",
+                "model.layers.10.self_attn.q_proj",
+                "model.layers.10.self_attn.qkv_proj",
+                "model.layers.10.self_attn.v_conv1d",
+                "model.layers.10.self_attn.v_proj",
+                "model.layers.11.hc_attn_base",
+                "model.layers.11.hc_attn_fn",
+                "model.layers.11.hc_attn_scale",
+                "model.layers.11.hc_ffn_base",
+                "model.layers.11.hc_ffn_fn",
+                "model.layers.11.hc_ffn_scale",
+                "model.layers.11.input_layernorm",
+                "model.layers.11.mlp.gate",
+                "model.layers.11.mlp.gate.e_score_correction_bias",
+                "model.layers.11.post_attention_layernorm",
+                "model.layers.11.self_attn.indexer.index_kpool_compress_ape",
+                "model.layers.11.self_attn.indexer.index_kpool_compress_gate",
+                "model.layers.11.self_attn.indexer.k_norm",
+                "model.layers.11.self_attn.indexer.k_norm.bias",
+                "model.layers.11.self_attn.indexer.weights_proj",
+                "model.layers.11.self_attn.indexer.wk",
+                "model.layers.11.self_attn.indexer.wq_b",
+                "model.layers.11.self_attn.kv_a_layernorm",
+                "model.layers.11.self_attn.kv_b_proj",
+                "model.layers.11.self_attn.q_a_layernorm",
+                "model.layers.12.hc_attn_base",
+                "model.layers.12.hc_attn_fn",
+                "model.layers.12.hc_attn_scale",
+                "model.layers.12.hc_ffn_base",
+                "model.layers.12.hc_ffn_fn",
+                "model.layers.12.hc_ffn_scale",
+                "model.layers.12.input_layernorm",
+                "model.layers.12.mlp.gate",
+                "model.layers.12.mlp.gate.e_score_correction_bias",
+                "model.layers.12.post_attention_layernorm",
+                "model.layers.12.self_attn.A_log",
+                "model.layers.12.self_attn.b_proj",
+                "model.layers.12.self_attn.dt_bias",
+                "model.layers.12.self_attn.f_a_proj",
+                "model.layers.12.self_attn.f_b_proj",
+                "model.layers.12.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.12.self_attn.g_a_proj",
+                "model.layers.12.self_attn.g_b_proj",
+                "model.layers.12.self_attn.k_conv1d",
+                "model.layers.12.self_attn.k_proj",
+                "model.layers.12.self_attn.o_norm",
+                "model.layers.12.self_attn.o_proj",
+                "model.layers.12.self_attn.q_conv1d",
+                "model.layers.12.self_attn.q_proj",
+                "model.layers.12.self_attn.qkv_proj",
+                "model.layers.12.self_attn.v_conv1d",
+                "model.layers.12.self_attn.v_proj",
+                "model.layers.13.hc_attn_base",
+                "model.layers.13.hc_attn_fn",
+                "model.layers.13.hc_attn_scale",
+                "model.layers.13.hc_ffn_base",
+                "model.layers.13.hc_ffn_fn",
+                "model.layers.13.hc_ffn_scale",
+                "model.layers.13.input_layernorm",
+                "model.layers.13.mlp.gate",
+                "model.layers.13.mlp.gate.e_score_correction_bias",
+                "model.layers.13.post_attention_layernorm",
+                "model.layers.13.self_attn.A_log",
+                "model.layers.13.self_attn.b_proj",
+                "model.layers.13.self_attn.dt_bias",
+                "model.layers.13.self_attn.f_a_proj",
+                "model.layers.13.self_attn.f_b_proj",
+                "model.layers.13.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.13.self_attn.g_a_proj",
+                "model.layers.13.self_attn.g_b_proj",
+                "model.layers.13.self_attn.k_conv1d",
+                "model.layers.13.self_attn.k_proj",
+                "model.layers.13.self_attn.o_norm",
+                "model.layers.13.self_attn.o_proj",
+                "model.layers.13.self_attn.q_conv1d",
+                "model.layers.13.self_attn.q_proj",
+                "model.layers.13.self_attn.qkv_proj",
+                "model.layers.13.self_attn.v_conv1d",
+                "model.layers.13.self_attn.v_proj",
+                "model.layers.14.hc_attn_base",
+                "model.layers.14.hc_attn_fn",
+                "model.layers.14.hc_attn_scale",
+                "model.layers.14.hc_ffn_base",
+                "model.layers.14.hc_ffn_fn",
+                "model.layers.14.hc_ffn_scale",
+                "model.layers.14.input_layernorm",
+                "model.layers.14.mlp.gate",
+                "model.layers.14.mlp.gate.e_score_correction_bias",
+                "model.layers.14.post_attention_layernorm",
+                "model.layers.14.self_attn.A_log",
+                "model.layers.14.self_attn.b_proj",
+                "model.layers.14.self_attn.dt_bias",
+                "model.layers.14.self_attn.f_a_proj",
+                "model.layers.14.self_attn.f_b_proj",
+                "model.layers.14.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.14.self_attn.g_a_proj",
+                "model.layers.14.self_attn.g_b_proj",
+                "model.layers.14.self_attn.k_conv1d",
+                "model.layers.14.self_attn.k_proj",
+                "model.layers.14.self_attn.o_norm",
+                "model.layers.14.self_attn.o_proj",
+                "model.layers.14.self_attn.q_conv1d",
+                "model.layers.14.self_attn.q_proj",
+                "model.layers.14.self_attn.qkv_proj",
+                "model.layers.14.self_attn.v_conv1d",
+                "model.layers.14.self_attn.v_proj",
+                "model.layers.15.hc_attn_base",
+                "model.layers.15.hc_attn_fn",
+                "model.layers.15.hc_attn_scale",
+                "model.layers.15.hc_ffn_base",
+                "model.layers.15.hc_ffn_fn",
+                "model.layers.15.hc_ffn_scale",
+                "model.layers.15.input_layernorm",
+                "model.layers.15.mlp.gate",
+                "model.layers.15.mlp.gate.e_score_correction_bias",
+                "model.layers.15.post_attention_layernorm",
+                "model.layers.15.self_attn.indexer.index_kpool_compress_ape",
+                "model.layers.15.self_attn.indexer.index_kpool_compress_gate",
+                "model.layers.15.self_attn.indexer.k_norm",
+                "model.layers.15.self_attn.indexer.k_norm.bias",
+                "model.layers.15.self_attn.indexer.weights_proj",
+                "model.layers.15.self_attn.indexer.wk",
+                "model.layers.15.self_attn.indexer.wq_b",
+                "model.layers.15.self_attn.kv_a_layernorm",
+                "model.layers.15.self_attn.kv_b_proj",
+                "model.layers.15.self_attn.q_a_layernorm",
+                "model.layers.16.hc_attn_base",
+                "model.layers.16.hc_attn_fn",
+                "model.layers.16.hc_attn_scale",
+                "model.layers.16.hc_ffn_base",
+                "model.layers.16.hc_ffn_fn",
+                "model.layers.16.hc_ffn_scale",
+                "model.layers.16.input_layernorm",
+                "model.layers.16.mlp.gate",
+                "model.layers.16.mlp.gate.e_score_correction_bias",
+                "model.layers.16.post_attention_layernorm",
+                "model.layers.16.self_attn.A_log",
+                "model.layers.16.self_attn.b_proj",
+                "model.layers.16.self_attn.dt_bias",
+                "model.layers.16.self_attn.f_a_proj",
+                "model.layers.16.self_attn.f_b_proj",
+                "model.layers.16.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.16.self_attn.g_a_proj",
+                "model.layers.16.self_attn.g_b_proj",
+                "model.layers.16.self_attn.k_conv1d",
+                "model.layers.16.self_attn.k_proj",
+                "model.layers.16.self_attn.o_norm",
+                "model.layers.16.self_attn.o_proj",
+                "model.layers.16.self_attn.q_conv1d",
+                "model.layers.16.self_attn.q_proj",
+                "model.layers.16.self_attn.qkv_proj",
+                "model.layers.16.self_attn.v_conv1d",
+                "model.layers.16.self_attn.v_proj",
+                "model.layers.17.hc_attn_base",
+                "model.layers.17.hc_attn_fn",
+                "model.layers.17.hc_attn_scale",
+                "model.layers.17.hc_ffn_base",
+                "model.layers.17.hc_ffn_fn",
+                "model.layers.17.hc_ffn_scale",
+                "model.layers.17.input_layernorm",
+                "model.layers.17.mlp.gate",
+                "model.layers.17.mlp.gate.e_score_correction_bias",
+                "model.layers.17.post_attention_layernorm",
+                "model.layers.17.self_attn.A_log",
+                "model.layers.17.self_attn.b_proj",
+                "model.layers.17.self_attn.dt_bias",
+                "model.layers.17.self_attn.f_a_proj",
+                "model.layers.17.self_attn.f_b_proj",
+                "model.layers.17.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.17.self_attn.g_a_proj",
+                "model.layers.17.self_attn.g_b_proj",
+                "model.layers.17.self_attn.k_conv1d",
+                "model.layers.17.self_attn.k_proj",
+                "model.layers.17.self_attn.o_norm",
+                "model.layers.17.self_attn.o_proj",
+                "model.layers.17.self_attn.q_conv1d",
+                "model.layers.17.self_attn.q_proj",
+                "model.layers.17.self_attn.qkv_proj",
+                "model.layers.17.self_attn.v_conv1d",
+                "model.layers.17.self_attn.v_proj",
+                "model.layers.18.hc_attn_base",
+                "model.layers.18.hc_attn_fn",
+                "model.layers.18.hc_attn_scale",
+                "model.layers.18.hc_ffn_base",
+                "model.layers.18.hc_ffn_fn",
+                "model.layers.18.hc_ffn_scale",
+                "model.layers.18.input_layernorm",
+                "model.layers.18.mlp.gate",
+                "model.layers.18.mlp.gate.e_score_correction_bias",
+                "model.layers.18.post_attention_layernorm",
+                "model.layers.18.self_attn.A_log",
+                "model.layers.18.self_attn.b_proj",
+                "model.layers.18.self_attn.dt_bias",
+                "model.layers.18.self_attn.f_a_proj",
+                "model.layers.18.self_attn.f_b_proj",
+                "model.layers.18.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.18.self_attn.g_a_proj",
+                "model.layers.18.self_attn.g_b_proj",
+                "model.layers.18.self_attn.k_conv1d",
+                "model.layers.18.self_attn.k_proj",
+                "model.layers.18.self_attn.o_norm",
+                "model.layers.18.self_attn.o_proj",
+                "model.layers.18.self_attn.q_conv1d",
+                "model.layers.18.self_attn.q_proj",
+                "model.layers.18.self_attn.qkv_proj",
+                "model.layers.18.self_attn.v_conv1d",
+                "model.layers.18.self_attn.v_proj",
+                "model.layers.19.hc_attn_base",
+                "model.layers.19.hc_attn_fn",
+                "model.layers.19.hc_attn_scale",
+                "model.layers.19.hc_ffn_base",
+                "model.layers.19.hc_ffn_fn",
+                "model.layers.19.hc_ffn_scale",
+                "model.layers.19.input_layernorm",
+                "model.layers.19.mlp.gate",
+                "model.layers.19.mlp.gate.e_score_correction_bias",
+                "model.layers.19.post_attention_layernorm",
+                "model.layers.19.self_attn.indexer.index_kpool_compress_ape",
+                "model.layers.19.self_attn.indexer.index_kpool_compress_gate",
+                "model.layers.19.self_attn.indexer.k_norm",
+                "model.layers.19.self_attn.indexer.k_norm.bias",
+                "model.layers.19.self_attn.indexer.weights_proj",
+                "model.layers.19.self_attn.indexer.wk",
+                "model.layers.19.self_attn.indexer.wq_b",
+                "model.layers.19.self_attn.kv_a_layernorm",
+                "model.layers.19.self_attn.kv_b_proj",
+                "model.layers.19.self_attn.q_a_layernorm",
+                "model.layers.2.hc_attn_base",
+                "model.layers.2.hc_attn_fn",
+                "model.layers.2.hc_attn_scale",
+                "model.layers.2.hc_ffn_base",
+                "model.layers.2.hc_ffn_fn",
+                "model.layers.2.hc_ffn_scale",
+                "model.layers.2.input_layernorm",
+                "model.layers.2.post_attention_layernorm",
+                "model.layers.2.self_attn.A_log",
+                "model.layers.2.self_attn.b_proj",
+                "model.layers.2.self_attn.dt_bias",
+                "model.layers.2.self_attn.f_a_proj",
+                "model.layers.2.self_attn.f_b_proj",
+                "model.layers.2.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.2.self_attn.g_a_proj",
+                "model.layers.2.self_attn.g_b_proj",
+                "model.layers.2.self_attn.k_conv1d",
+                "model.layers.2.self_attn.k_proj",
+                "model.layers.2.self_attn.o_norm",
+                "model.layers.2.self_attn.o_proj",
+                "model.layers.2.self_attn.q_conv1d",
+                "model.layers.2.self_attn.q_proj",
+                "model.layers.2.self_attn.qkv_proj",
+                "model.layers.2.self_attn.v_conv1d",
+                "model.layers.2.self_attn.v_proj",
+                "model.layers.20.hc_attn_base",
+                "model.layers.20.hc_attn_fn",
+                "model.layers.20.hc_attn_scale",
+                "model.layers.20.hc_ffn_base",
+                "model.layers.20.hc_ffn_fn",
+                "model.layers.20.hc_ffn_scale",
+                "model.layers.20.input_layernorm",
+                "model.layers.20.mlp.gate",
+                "model.layers.20.mlp.gate.e_score_correction_bias",
+                "model.layers.20.post_attention_layernorm",
+                "model.layers.20.self_attn.A_log",
+                "model.layers.20.self_attn.b_proj",
+                "model.layers.20.self_attn.dt_bias",
+                "model.layers.20.self_attn.f_a_proj",
+                "model.layers.20.self_attn.f_b_proj",
+                "model.layers.20.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.20.self_attn.g_a_proj",
+                "model.layers.20.self_attn.g_b_proj",
+                "model.layers.20.self_attn.k_conv1d",
+                "model.layers.20.self_attn.k_proj",
+                "model.layers.20.self_attn.o_norm",
+                "model.layers.20.self_attn.o_proj",
+                "model.layers.20.self_attn.q_conv1d",
+                "model.layers.20.self_attn.q_proj",
+                "model.layers.20.self_attn.qkv_proj",
+                "model.layers.20.self_attn.v_conv1d",
+                "model.layers.20.self_attn.v_proj",
+                "model.layers.21.hc_attn_base",
+                "model.layers.21.hc_attn_fn",
+                "model.layers.21.hc_attn_scale",
+                "model.layers.21.hc_ffn_base",
+                "model.layers.21.hc_ffn_fn",
+                "model.layers.21.hc_ffn_scale",
+                "model.layers.21.input_layernorm",
+                "model.layers.21.mlp.gate",
+                "model.layers.21.mlp.gate.e_score_correction_bias",
+                "model.layers.21.post_attention_layernorm",
+                "model.layers.21.self_attn.A_log",
+                "model.layers.21.self_attn.b_proj",
+                "model.layers.21.self_attn.dt_bias",
+                "model.layers.21.self_attn.f_a_proj",
+                "model.layers.21.self_attn.f_b_proj",
+                "model.layers.21.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.21.self_attn.g_a_proj",
+                "model.layers.21.self_attn.g_b_proj",
+                "model.layers.21.self_attn.k_conv1d",
+                "model.layers.21.self_attn.k_proj",
+                "model.layers.21.self_attn.o_norm",
+                "model.layers.21.self_attn.o_proj",
+                "model.layers.21.self_attn.q_conv1d",
+                "model.layers.21.self_attn.q_proj",
+                "model.layers.21.self_attn.qkv_proj",
+                "model.layers.21.self_attn.v_conv1d",
+                "model.layers.21.self_attn.v_proj",
+                "model.layers.22.hc_attn_base",
+                "model.layers.22.hc_attn_fn",
+                "model.layers.22.hc_attn_scale",
+                "model.layers.22.hc_ffn_base",
+                "model.layers.22.hc_ffn_fn",
+                "model.layers.22.hc_ffn_scale",
+                "model.layers.22.input_layernorm",
+                "model.layers.22.mlp.gate",
+                "model.layers.22.mlp.gate.e_score_correction_bias",
+                "model.layers.22.post_attention_layernorm",
+                "model.layers.22.self_attn.A_log",
+                "model.layers.22.self_attn.b_proj",
+                "model.layers.22.self_attn.dt_bias",
+                "model.layers.22.self_attn.f_a_proj",
+                "model.layers.22.self_attn.f_b_proj",
+                "model.layers.22.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.22.self_attn.g_a_proj",
+                "model.layers.22.self_attn.g_b_proj",
+                "model.layers.22.self_attn.k_conv1d",
+                "model.layers.22.self_attn.k_proj",
+                "model.layers.22.self_attn.o_norm",
+                "model.layers.22.self_attn.o_proj",
+                "model.layers.22.self_attn.q_conv1d",
+                "model.layers.22.self_attn.q_proj",
+                "model.layers.22.self_attn.qkv_proj",
+                "model.layers.22.self_attn.v_conv1d",
+                "model.layers.22.self_attn.v_proj",
+                "model.layers.23.hc_attn_base",
+                "model.layers.23.hc_attn_fn",
+                "model.layers.23.hc_attn_scale",
+                "model.layers.23.hc_ffn_base",
+                "model.layers.23.hc_ffn_fn",
+                "model.layers.23.hc_ffn_scale",
+                "model.layers.23.input_layernorm",
+                "model.layers.23.mlp.gate",
+                "model.layers.23.mlp.gate.e_score_correction_bias",
+                "model.layers.23.post_attention_layernorm",
+                "model.layers.23.self_attn.indexer.index_kpool_compress_ape",
+                "model.layers.23.self_attn.indexer.index_kpool_compress_gate",
+                "model.layers.23.self_attn.indexer.k_norm",
+                "model.layers.23.self_attn.indexer.k_norm.bias",
+                "model.layers.23.self_attn.indexer.weights_proj",
+                "model.layers.23.self_attn.indexer.wk",
+                "model.layers.23.self_attn.indexer.wq_b",
+                "model.layers.23.self_attn.kv_a_layernorm",
+                "model.layers.23.self_attn.kv_b_proj",
+                "model.layers.23.self_attn.q_a_layernorm",
+                "model.layers.24.hc_attn_base",
+                "model.layers.24.hc_attn_fn",
+                "model.layers.24.hc_attn_scale",
+                "model.layers.24.hc_ffn_base",
+                "model.layers.24.hc_ffn_fn",
+                "model.layers.24.hc_ffn_scale",
+                "model.layers.24.input_layernorm",
+                "model.layers.24.mlp.gate",
+                "model.layers.24.mlp.gate.e_score_correction_bias",
+                "model.layers.24.post_attention_layernorm",
+                "model.layers.24.self_attn.A_log",
+                "model.layers.24.self_attn.b_proj",
+                "model.layers.24.self_attn.dt_bias",
+                "model.layers.24.self_attn.f_a_proj",
+                "model.layers.24.self_attn.f_b_proj",
+                "model.layers.24.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.24.self_attn.g_a_proj",
+                "model.layers.24.self_attn.g_b_proj",
+                "model.layers.24.self_attn.k_conv1d",
+                "model.layers.24.self_attn.k_proj",
+                "model.layers.24.self_attn.o_norm",
+                "model.layers.24.self_attn.o_proj",
+                "model.layers.24.self_attn.q_conv1d",
+                "model.layers.24.self_attn.q_proj",
+                "model.layers.24.self_attn.qkv_proj",
+                "model.layers.24.self_attn.v_conv1d",
+                "model.layers.24.self_attn.v_proj",
+                "model.layers.25.hc_attn_base",
+                "model.layers.25.hc_attn_fn",
+                "model.layers.25.hc_attn_scale",
+                "model.layers.25.hc_ffn_base",
+                "model.layers.25.hc_ffn_fn",
+                "model.layers.25.hc_ffn_scale",
+                "model.layers.25.input_layernorm",
+                "model.layers.25.mlp.gate",
+                "model.layers.25.mlp.gate.e_score_correction_bias",
+                "model.layers.25.post_attention_layernorm",
+                "model.layers.25.self_attn.A_log",
+                "model.layers.25.self_attn.b_proj",
+                "model.layers.25.self_attn.dt_bias",
+                "model.layers.25.self_attn.f_a_proj",
+                "model.layers.25.self_attn.f_b_proj",
+                "model.layers.25.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.25.self_attn.g_a_proj",
+                "model.layers.25.self_attn.g_b_proj",
+                "model.layers.25.self_attn.k_conv1d",
+                "model.layers.25.self_attn.k_proj",
+                "model.layers.25.self_attn.o_norm",
+                "model.layers.25.self_attn.o_proj",
+                "model.layers.25.self_attn.q_conv1d",
+                "model.layers.25.self_attn.q_proj",
+                "model.layers.25.self_attn.qkv_proj",
+                "model.layers.25.self_attn.v_conv1d",
+                "model.layers.25.self_attn.v_proj",
+                "model.layers.26.hc_attn_base",
+                "model.layers.26.hc_attn_fn",
+                "model.layers.26.hc_attn_scale",
+                "model.layers.26.hc_ffn_base",
+                "model.layers.26.hc_ffn_fn",
+                "model.layers.26.hc_ffn_scale",
+                "model.layers.26.input_layernorm",
+                "model.layers.26.mlp.gate",
+                "model.layers.26.mlp.gate.e_score_correction_bias",
+                "model.layers.26.post_attention_layernorm",
+                "model.layers.26.self_attn.A_log",
+                "model.layers.26.self_attn.b_proj",
+                "model.layers.26.self_attn.dt_bias",
+                "model.layers.26.self_attn.f_a_proj",
+                "model.layers.26.self_attn.f_b_proj",
+                "model.layers.26.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.26.self_attn.g_a_proj",
+                "model.layers.26.self_attn.g_b_proj",
+                "model.layers.26.self_attn.k_conv1d",
+                "model.layers.26.self_attn.k_proj",
+                "model.layers.26.self_attn.o_norm",
+                "model.layers.26.self_attn.o_proj",
+                "model.layers.26.self_attn.q_conv1d",
+                "model.layers.26.self_attn.q_proj",
+                "model.layers.26.self_attn.qkv_proj",
+                "model.layers.26.self_attn.v_conv1d",
+                "model.layers.26.self_attn.v_proj",
+                "model.layers.27.hc_attn_base",
+                "model.layers.27.hc_attn_fn",
+                "model.layers.27.hc_attn_scale",
+                "model.layers.27.hc_ffn_base",
+                "model.layers.27.hc_ffn_fn",
+                "model.layers.27.hc_ffn_scale",
+                "model.layers.27.input_layernorm",
+                "model.layers.27.mlp.gate",
+                "model.layers.27.mlp.gate.e_score_correction_bias",
+                "model.layers.27.post_attention_layernorm",
+                "model.layers.27.self_attn.indexer.index_kpool_compress_ape",
+                "model.layers.27.self_attn.indexer.index_kpool_compress_gate",
+                "model.layers.27.self_attn.indexer.k_norm",
+                "model.layers.27.self_attn.indexer.k_norm.bias",
+                "model.layers.27.self_attn.indexer.weights_proj",
+                "model.layers.27.self_attn.indexer.wk",
+                "model.layers.27.self_attn.indexer.wq_b",
+                "model.layers.27.self_attn.kv_a_layernorm",
+                "model.layers.27.self_attn.kv_b_proj",
+                "model.layers.27.self_attn.q_a_layernorm",
+                "model.layers.28.hc_attn_base",
+                "model.layers.28.hc_attn_fn",
+                "model.layers.28.hc_attn_scale",
+                "model.layers.28.hc_ffn_base",
+                "model.layers.28.hc_ffn_fn",
+                "model.layers.28.hc_ffn_scale",
+                "model.layers.28.input_layernorm",
+                "model.layers.28.mlp.gate",
+                "model.layers.28.mlp.gate.e_score_correction_bias",
+                "model.layers.28.post_attention_layernorm",
+                "model.layers.28.self_attn.A_log",
+                "model.layers.28.self_attn.b_proj",
+                "model.layers.28.self_attn.dt_bias",
+                "model.layers.28.self_attn.f_a_proj",
+                "model.layers.28.self_attn.f_b_proj",
+                "model.layers.28.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.28.self_attn.g_a_proj",
+                "model.layers.28.self_attn.g_b_proj",
+                "model.layers.28.self_attn.k_conv1d",
+                "model.layers.28.self_attn.k_proj",
+                "model.layers.28.self_attn.o_norm",
+                "model.layers.28.self_attn.o_proj",
+                "model.layers.28.self_attn.q_conv1d",
+                "model.layers.28.self_attn.q_proj",
+                "model.layers.28.self_attn.qkv_proj",
+                "model.layers.28.self_attn.v_conv1d",
+                "model.layers.28.self_attn.v_proj",
+                "model.layers.29.hc_attn_base",
+                "model.layers.29.hc_attn_fn",
+                "model.layers.29.hc_attn_scale",
+                "model.layers.29.hc_ffn_base",
+                "model.layers.29.hc_ffn_fn",
+                "model.layers.29.hc_ffn_scale",
+                "model.layers.29.input_layernorm",
+                "model.layers.29.mlp.gate",
+                "model.layers.29.mlp.gate.e_score_correction_bias",
+                "model.layers.29.post_attention_layernorm",
+                "model.layers.29.self_attn.A_log",
+                "model.layers.29.self_attn.b_proj",
+                "model.layers.29.self_attn.dt_bias",
+                "model.layers.29.self_attn.f_a_proj",
+                "model.layers.29.self_attn.f_b_proj",
+                "model.layers.29.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.29.self_attn.g_a_proj",
+                "model.layers.29.self_attn.g_b_proj",
+                "model.layers.29.self_attn.k_conv1d",
+                "model.layers.29.self_attn.k_proj",
+                "model.layers.29.self_attn.o_norm",
+                "model.layers.29.self_attn.o_proj",
+                "model.layers.29.self_attn.q_conv1d",
+                "model.layers.29.self_attn.q_proj",
+                "model.layers.29.self_attn.qkv_proj",
+                "model.layers.29.self_attn.v_conv1d",
+                "model.layers.29.self_attn.v_proj",
+                "model.layers.3.hc_attn_base",
+                "model.layers.3.hc_attn_fn",
+                "model.layers.3.hc_attn_scale",
+                "model.layers.3.hc_ffn_base",
+                "model.layers.3.hc_ffn_fn",
+                "model.layers.3.hc_ffn_scale",
+                "model.layers.3.input_layernorm",
+                "model.layers.3.mlp.gate",
+                "model.layers.3.mlp.gate.e_score_correction_bias",
+                "model.layers.3.post_attention_layernorm",
+                "model.layers.3.self_attn.indexer.index_kpool_compress_ape",
+                "model.layers.3.self_attn.indexer.index_kpool_compress_gate",
+                "model.layers.3.self_attn.indexer.k_norm",
+                "model.layers.3.self_attn.indexer.k_norm.bias",
+                "model.layers.3.self_attn.indexer.weights_proj",
+                "model.layers.3.self_attn.indexer.wk",
+                "model.layers.3.self_attn.indexer.wq_b",
+                "model.layers.3.self_attn.kv_a_layernorm",
+                "model.layers.3.self_attn.kv_b_proj",
+                "model.layers.3.self_attn.q_a_layernorm",
+                "model.layers.30.hc_attn_base",
+                "model.layers.30.hc_attn_fn",
+                "model.layers.30.hc_attn_scale",
+                "model.layers.30.hc_ffn_base",
+                "model.layers.30.hc_ffn_fn",
+                "model.layers.30.hc_ffn_scale",
+                "model.layers.30.input_layernorm",
+                "model.layers.30.mlp.gate",
+                "model.layers.30.mlp.gate.e_score_correction_bias",
+                "model.layers.30.post_attention_layernorm",
+                "model.layers.30.self_attn.A_log",
+                "model.layers.30.self_attn.b_proj",
+                "model.layers.30.self_attn.dt_bias",
+                "model.layers.30.self_attn.f_a_proj",
+                "model.layers.30.self_attn.f_b_proj",
+                "model.layers.30.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.30.self_attn.g_a_proj",
+                "model.layers.30.self_attn.g_b_proj",
+                "model.layers.30.self_attn.k_conv1d",
+                "model.layers.30.self_attn.k_proj",
+                "model.layers.30.self_attn.o_norm",
+                "model.layers.30.self_attn.o_proj",
+                "model.layers.30.self_attn.q_conv1d",
+                "model.layers.30.self_attn.q_proj",
+                "model.layers.30.self_attn.qkv_proj",
+                "model.layers.30.self_attn.v_conv1d",
+                "model.layers.30.self_attn.v_proj",
+                "model.layers.31.hc_attn_base",
+                "model.layers.31.hc_attn_fn",
+                "model.layers.31.hc_attn_scale",
+                "model.layers.31.hc_ffn_base",
+                "model.layers.31.hc_ffn_fn",
+                "model.layers.31.hc_ffn_scale",
+                "model.layers.31.input_layernorm",
+                "model.layers.31.mlp.gate",
+                "model.layers.31.mlp.gate.e_score_correction_bias",
+                "model.layers.31.post_attention_layernorm",
+                "model.layers.31.self_attn.indexer.index_kpool_compress_ape",
+                "model.layers.31.self_attn.indexer.index_kpool_compress_gate",
+                "model.layers.31.self_attn.indexer.k_norm",
+                "model.layers.31.self_attn.indexer.k_norm.bias",
+                "model.layers.31.self_attn.indexer.weights_proj",
+                "model.layers.31.self_attn.indexer.wk",
+                "model.layers.31.self_attn.indexer.wq_b",
+                "model.layers.31.self_attn.kv_a_layernorm",
+                "model.layers.31.self_attn.kv_b_proj",
+                "model.layers.31.self_attn.q_a_layernorm",
+                "model.layers.32.hc_attn_base",
+                "model.layers.32.hc_attn_fn",
+                "model.layers.32.hc_attn_scale",
+                "model.layers.32.hc_ffn_base",
+                "model.layers.32.hc_ffn_fn",
+                "model.layers.32.hc_ffn_scale",
+                "model.layers.32.input_layernorm",
+                "model.layers.32.mlp.gate",
+                "model.layers.32.mlp.gate.e_score_correction_bias",
+                "model.layers.32.post_attention_layernorm",
+                "model.layers.32.self_attn.A_log",
+                "model.layers.32.self_attn.b_proj",
+                "model.layers.32.self_attn.dt_bias",
+                "model.layers.32.self_attn.f_a_proj",
+                "model.layers.32.self_attn.f_b_proj",
+                "model.layers.32.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.32.self_attn.g_a_proj",
+                "model.layers.32.self_attn.g_b_proj",
+                "model.layers.32.self_attn.k_conv1d",
+                "model.layers.32.self_attn.k_proj",
+                "model.layers.32.self_attn.o_norm",
+                "model.layers.32.self_attn.o_proj",
+                "model.layers.32.self_attn.q_conv1d",
+                "model.layers.32.self_attn.q_proj",
+                "model.layers.32.self_attn.qkv_proj",
+                "model.layers.32.self_attn.v_conv1d",
+                "model.layers.32.self_attn.v_proj",
+                "model.layers.33.hc_attn_base",
+                "model.layers.33.hc_attn_fn",
+                "model.layers.33.hc_attn_scale",
+                "model.layers.33.hc_ffn_base",
+                "model.layers.33.hc_ffn_fn",
+                "model.layers.33.hc_ffn_scale",
+                "model.layers.33.input_layernorm",
+                "model.layers.33.mlp.gate",
+                "model.layers.33.mlp.gate.e_score_correction_bias",
+                "model.layers.33.post_attention_layernorm",
+                "model.layers.33.self_attn.A_log",
+                "model.layers.33.self_attn.b_proj",
+                "model.layers.33.self_attn.dt_bias",
+                "model.layers.33.self_attn.f_a_proj",
+                "model.layers.33.self_attn.f_b_proj",
+                "model.layers.33.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.33.self_attn.g_a_proj",
+                "model.layers.33.self_attn.g_b_proj",
+                "model.layers.33.self_attn.k_conv1d",
+                "model.layers.33.self_attn.k_proj",
+                "model.layers.33.self_attn.o_norm",
+                "model.layers.33.self_attn.o_proj",
+                "model.layers.33.self_attn.q_conv1d",
+                "model.layers.33.self_attn.q_proj",
+                "model.layers.33.self_attn.qkv_proj",
+                "model.layers.33.self_attn.v_conv1d",
+                "model.layers.33.self_attn.v_proj",
+                "model.layers.34.hc_attn_base",
+                "model.layers.34.hc_attn_fn",
+                "model.layers.34.hc_attn_scale",
+                "model.layers.34.hc_ffn_base",
+                "model.layers.34.hc_ffn_fn",
+                "model.layers.34.hc_ffn_scale",
+                "model.layers.34.input_layernorm",
+                "model.layers.34.mlp.gate",
+                "model.layers.34.mlp.gate.e_score_correction_bias",
+                "model.layers.34.post_attention_layernorm",
+                "model.layers.34.self_attn.A_log",
+                "model.layers.34.self_attn.b_proj",
+                "model.layers.34.self_attn.dt_bias",
+                "model.layers.34.self_attn.f_a_proj",
+                "model.layers.34.self_attn.f_b_proj",
+                "model.layers.34.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.34.self_attn.g_a_proj",
+                "model.layers.34.self_attn.g_b_proj",
+                "model.layers.34.self_attn.k_conv1d",
+                "model.layers.34.self_attn.k_proj",
+                "model.layers.34.self_attn.o_norm",
+                "model.layers.34.self_attn.o_proj",
+                "model.layers.34.self_attn.q_conv1d",
+                "model.layers.34.self_attn.q_proj",
+                "model.layers.34.self_attn.qkv_proj",
+                "model.layers.34.self_attn.v_conv1d",
+                "model.layers.34.self_attn.v_proj",
+                "model.layers.35.hc_attn_base",
+                "model.layers.35.hc_attn_fn",
+                "model.layers.35.hc_attn_scale",
+                "model.layers.35.hc_ffn_base",
+                "model.layers.35.hc_ffn_fn",
+                "model.layers.35.hc_ffn_scale",
+                "model.layers.35.input_layernorm",
+                "model.layers.35.mlp.gate",
+                "model.layers.35.mlp.gate.e_score_correction_bias",
+                "model.layers.35.post_attention_layernorm",
+                "model.layers.35.self_attn.indexer.index_kpool_compress_ape",
+                "model.layers.35.self_attn.indexer.index_kpool_compress_gate",
+                "model.layers.35.self_attn.indexer.k_norm",
+                "model.layers.35.self_attn.indexer.k_norm.bias",
+                "model.layers.35.self_attn.indexer.weights_proj",
+                "model.layers.35.self_attn.indexer.wk",
+                "model.layers.35.self_attn.indexer.wq_b",
+                "model.layers.35.self_attn.kv_a_layernorm",
+                "model.layers.35.self_attn.kv_b_proj",
+                "model.layers.35.self_attn.q_a_layernorm",
+                "model.layers.36.hc_attn_base",
+                "model.layers.36.hc_attn_fn",
+                "model.layers.36.hc_attn_scale",
+                "model.layers.36.hc_ffn_base",
+                "model.layers.36.hc_ffn_fn",
+                "model.layers.36.hc_ffn_scale",
+                "model.layers.36.input_layernorm",
+                "model.layers.36.mlp.gate",
+                "model.layers.36.mlp.gate.e_score_correction_bias",
+                "model.layers.36.post_attention_layernorm",
+                "model.layers.36.self_attn.A_log",
+                "model.layers.36.self_attn.b_proj",
+                "model.layers.36.self_attn.dt_bias",
+                "model.layers.36.self_attn.f_a_proj",
+                "model.layers.36.self_attn.f_b_proj",
+                "model.layers.36.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.36.self_attn.g_a_proj",
+                "model.layers.36.self_attn.g_b_proj",
+                "model.layers.36.self_attn.k_conv1d",
+                "model.layers.36.self_attn.k_proj",
+                "model.layers.36.self_attn.o_norm",
+                "model.layers.36.self_attn.o_proj",
+                "model.layers.36.self_attn.q_conv1d",
+                "model.layers.36.self_attn.q_proj",
+                "model.layers.36.self_attn.qkv_proj",
+                "model.layers.36.self_attn.v_conv1d",
+                "model.layers.36.self_attn.v_proj",
+                "model.layers.37.hc_attn_base",
+                "model.layers.37.hc_attn_fn",
+                "model.layers.37.hc_attn_scale",
+                "model.layers.37.hc_ffn_base",
+                "model.layers.37.hc_ffn_fn",
+                "model.layers.37.hc_ffn_scale",
+                "model.layers.37.input_layernorm",
+                "model.layers.37.mlp.gate",
+                "model.layers.37.mlp.gate.e_score_correction_bias",
+                "model.layers.37.post_attention_layernorm",
+                "model.layers.37.self_attn.A_log",
+                "model.layers.37.self_attn.b_proj",
+                "model.layers.37.self_attn.dt_bias",
+                "model.layers.37.self_attn.f_a_proj",
+                "model.layers.37.self_attn.f_b_proj",
+                "model.layers.37.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.37.self_attn.g_a_proj",
+                "model.layers.37.self_attn.g_b_proj",
+                "model.layers.37.self_attn.k_conv1d",
+                "model.layers.37.self_attn.k_proj",
+                "model.layers.37.self_attn.o_norm",
+                "model.layers.37.self_attn.o_proj",
+                "model.layers.37.self_attn.q_conv1d",
+                "model.layers.37.self_attn.q_proj",
+                "model.layers.37.self_attn.qkv_proj",
+                "model.layers.37.self_attn.v_conv1d",
+                "model.layers.37.self_attn.v_proj",
+                "model.layers.38.hc_attn_base",
+                "model.layers.38.hc_attn_fn",
+                "model.layers.38.hc_attn_scale",
+                "model.layers.38.hc_ffn_base",
+                "model.layers.38.hc_ffn_fn",
+                "model.layers.38.hc_ffn_scale",
+                "model.layers.38.input_layernorm",
+                "model.layers.38.mlp.gate",
+                "model.layers.38.mlp.gate.e_score_correction_bias",
+                "model.layers.38.post_attention_layernorm",
+                "model.layers.38.self_attn.A_log",
+                "model.layers.38.self_attn.b_proj",
+                "model.layers.38.self_attn.dt_bias",
+                "model.layers.38.self_attn.f_a_proj",
+                "model.layers.38.self_attn.f_b_proj",
+                "model.layers.38.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.38.self_attn.g_a_proj",
+                "model.layers.38.self_attn.g_b_proj",
+                "model.layers.38.self_attn.k_conv1d",
+                "model.layers.38.self_attn.k_proj",
+                "model.layers.38.self_attn.o_norm",
+                "model.layers.38.self_attn.o_proj",
+                "model.layers.38.self_attn.q_conv1d",
+                "model.layers.38.self_attn.q_proj",
+                "model.layers.38.self_attn.qkv_proj",
+                "model.layers.38.self_attn.v_conv1d",
+                "model.layers.38.self_attn.v_proj",
+                "model.layers.39.hc_attn_base",
+                "model.layers.39.hc_attn_fn",
+                "model.layers.39.hc_attn_scale",
+                "model.layers.39.hc_ffn_base",
+                "model.layers.39.hc_ffn_fn",
+                "model.layers.39.hc_ffn_scale",
+                "model.layers.39.input_layernorm",
+                "model.layers.39.mlp.gate",
+                "model.layers.39.mlp.gate.e_score_correction_bias",
+                "model.layers.39.post_attention_layernorm",
+                "model.layers.39.self_attn.indexer.index_kpool_compress_ape",
+                "model.layers.39.self_attn.indexer.index_kpool_compress_gate",
+                "model.layers.39.self_attn.indexer.k_norm",
+                "model.layers.39.self_attn.indexer.k_norm.bias",
+                "model.layers.39.self_attn.indexer.weights_proj",
+                "model.layers.39.self_attn.indexer.wk",
+                "model.layers.39.self_attn.indexer.wq_b",
+                "model.layers.39.self_attn.kv_a_layernorm",
+                "model.layers.39.self_attn.kv_b_proj",
+                "model.layers.39.self_attn.q_a_layernorm",
+                "model.layers.4.hc_attn_base",
+                "model.layers.4.hc_attn_fn",
+                "model.layers.4.hc_attn_scale",
+                "model.layers.4.hc_ffn_base",
+                "model.layers.4.hc_ffn_fn",
+                "model.layers.4.hc_ffn_scale",
+                "model.layers.4.input_layernorm",
+                "model.layers.4.mlp.gate",
+                "model.layers.4.mlp.gate.e_score_correction_bias",
+                "model.layers.4.post_attention_layernorm",
+                "model.layers.4.self_attn.A_log",
+                "model.layers.4.self_attn.b_proj",
+                "model.layers.4.self_attn.dt_bias",
+                "model.layers.4.self_attn.f_a_proj",
+                "model.layers.4.self_attn.f_b_proj",
+                "model.layers.4.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.4.self_attn.g_a_proj",
+                "model.layers.4.self_attn.g_b_proj",
+                "model.layers.4.self_attn.k_conv1d",
+                "model.layers.4.self_attn.k_proj",
+                "model.layers.4.self_attn.o_norm",
+                "model.layers.4.self_attn.o_proj",
+                "model.layers.4.self_attn.q_conv1d",
+                "model.layers.4.self_attn.q_proj",
+                "model.layers.4.self_attn.qkv_proj",
+                "model.layers.4.self_attn.v_conv1d",
+                "model.layers.4.self_attn.v_proj",
+                "model.layers.40.hc_attn_base",
+                "model.layers.40.hc_attn_fn",
+                "model.layers.40.hc_attn_scale",
+                "model.layers.40.hc_ffn_base",
+                "model.layers.40.hc_ffn_fn",
+                "model.layers.40.hc_ffn_scale",
+                "model.layers.40.input_layernorm",
+                "model.layers.40.mlp.gate",
+                "model.layers.40.mlp.gate.e_score_correction_bias",
+                "model.layers.40.post_attention_layernorm",
+                "model.layers.40.self_attn.A_log",
+                "model.layers.40.self_attn.b_proj",
+                "model.layers.40.self_attn.dt_bias",
+                "model.layers.40.self_attn.f_a_proj",
+                "model.layers.40.self_attn.f_b_proj",
+                "model.layers.40.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.40.self_attn.g_a_proj",
+                "model.layers.40.self_attn.g_b_proj",
+                "model.layers.40.self_attn.k_conv1d",
+                "model.layers.40.self_attn.k_proj",
+                "model.layers.40.self_attn.o_norm",
+                "model.layers.40.self_attn.o_proj",
+                "model.layers.40.self_attn.q_conv1d",
+                "model.layers.40.self_attn.q_proj",
+                "model.layers.40.self_attn.qkv_proj",
+                "model.layers.40.self_attn.v_conv1d",
+                "model.layers.40.self_attn.v_proj",
+                "model.layers.41.hc_attn_base",
+                "model.layers.41.hc_attn_fn",
+                "model.layers.41.hc_attn_scale",
+                "model.layers.41.hc_ffn_base",
+                "model.layers.41.hc_ffn_fn",
+                "model.layers.41.hc_ffn_scale",
+                "model.layers.41.input_layernorm",
+                "model.layers.41.mlp.gate",
+                "model.layers.41.mlp.gate.e_score_correction_bias",
+                "model.layers.41.post_attention_layernorm",
+                "model.layers.41.self_attn.A_log",
+                "model.layers.41.self_attn.b_proj",
+                "model.layers.41.self_attn.dt_bias",
+                "model.layers.41.self_attn.f_a_proj",
+                "model.layers.41.self_attn.f_b_proj",
+                "model.layers.41.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.41.self_attn.g_a_proj",
+                "model.layers.41.self_attn.g_b_proj",
+                "model.layers.41.self_attn.k_conv1d",
+                "model.layers.41.self_attn.k_proj",
+                "model.layers.41.self_attn.o_norm",
+                "model.layers.41.self_attn.o_proj",
+                "model.layers.41.self_attn.q_conv1d",
+                "model.layers.41.self_attn.q_proj",
+                "model.layers.41.self_attn.qkv_proj",
+                "model.layers.41.self_attn.v_conv1d",
+                "model.layers.41.self_attn.v_proj",
+                "model.layers.42.hc_attn_base",
+                "model.layers.42.hc_attn_fn",
+                "model.layers.42.hc_attn_scale",
+                "model.layers.42.hc_ffn_base",
+                "model.layers.42.hc_ffn_fn",
+                "model.layers.42.hc_ffn_scale",
+                "model.layers.42.input_layernorm",
+                "model.layers.42.mlp.gate",
+                "model.layers.42.mlp.gate.e_score_correction_bias",
+                "model.layers.42.post_attention_layernorm",
+                "model.layers.42.self_attn.A_log",
+                "model.layers.42.self_attn.b_proj",
+                "model.layers.42.self_attn.dt_bias",
+                "model.layers.42.self_attn.f_a_proj",
+                "model.layers.42.self_attn.f_b_proj",
+                "model.layers.42.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.42.self_attn.g_a_proj",
+                "model.layers.42.self_attn.g_b_proj",
+                "model.layers.42.self_attn.k_conv1d",
+                "model.layers.42.self_attn.k_proj",
+                "model.layers.42.self_attn.o_norm",
+                "model.layers.42.self_attn.o_proj",
+                "model.layers.42.self_attn.q_conv1d",
+                "model.layers.42.self_attn.q_proj",
+                "model.layers.42.self_attn.qkv_proj",
+                "model.layers.42.self_attn.v_conv1d",
+                "model.layers.42.self_attn.v_proj",
+                "model.layers.43.hc_attn_base",
+                "model.layers.43.hc_attn_fn",
+                "model.layers.43.hc_attn_scale",
+                "model.layers.43.hc_ffn_base",
+                "model.layers.43.hc_ffn_fn",
+                "model.layers.43.hc_ffn_scale",
+                "model.layers.43.input_layernorm",
+                "model.layers.43.mlp.gate",
+                "model.layers.43.mlp.gate.e_score_correction_bias",
+                "model.layers.43.post_attention_layernorm",
+                "model.layers.43.self_attn.indexer.index_kpool_compress_ape",
+                "model.layers.43.self_attn.indexer.index_kpool_compress_gate",
+                "model.layers.43.self_attn.indexer.k_norm",
+                "model.layers.43.self_attn.indexer.k_norm.bias",
+                "model.layers.43.self_attn.indexer.weights_proj",
+                "model.layers.43.self_attn.indexer.wk",
+                "model.layers.43.self_attn.indexer.wq_b",
+                "model.layers.43.self_attn.kv_a_layernorm",
+                "model.layers.43.self_attn.kv_b_proj",
+                "model.layers.43.self_attn.q_a_layernorm",
+                "model.layers.44.hc_attn_base",
+                "model.layers.44.hc_attn_fn",
+                "model.layers.44.hc_attn_scale",
+                "model.layers.44.hc_ffn_base",
+                "model.layers.44.hc_ffn_fn",
+                "model.layers.44.hc_ffn_scale",
+                "model.layers.44.input_layernorm",
+                "model.layers.44.mlp.gate",
+                "model.layers.44.mlp.gate.e_score_correction_bias",
+                "model.layers.44.post_attention_layernorm",
+                "model.layers.44.self_attn.A_log",
+                "model.layers.44.self_attn.b_proj",
+                "model.layers.44.self_attn.dt_bias",
+                "model.layers.44.self_attn.f_a_proj",
+                "model.layers.44.self_attn.f_b_proj",
+                "model.layers.44.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.44.self_attn.g_a_proj",
+                "model.layers.44.self_attn.g_b_proj",
+                "model.layers.44.self_attn.k_conv1d",
+                "model.layers.44.self_attn.k_proj",
+                "model.layers.44.self_attn.o_norm",
+                "model.layers.44.self_attn.o_proj",
+                "model.layers.44.self_attn.q_conv1d",
+                "model.layers.44.self_attn.q_proj",
+                "model.layers.44.self_attn.qkv_proj",
+                "model.layers.44.self_attn.v_conv1d",
+                "model.layers.44.self_attn.v_proj",
+                "model.layers.45.eh_proj",
+                "model.layers.45.enorm",
+                "model.layers.45.hnorm",
+                "model.layers.45.input_layernorm",
+                "model.layers.45.mlp.gate",
+                "model.layers.45.mlp.gate.e_score_correction_bias",
+                "model.layers.45.post_attention_layernorm",
+                "model.layers.45.self_attn.indexer.index_kpool_compress_ape",
+                "model.layers.45.self_attn.indexer.index_kpool_compress_gate",
+                "model.layers.45.self_attn.indexer.k_norm",
+                "model.layers.45.self_attn.indexer.k_norm.bias",
+                "model.layers.45.self_attn.indexer.weights_proj",
+                "model.layers.45.self_attn.indexer.wk",
+                "model.layers.45.self_attn.indexer.wq_b",
+                "model.layers.45.self_attn.kv_a_layernorm",
+                "model.layers.45.self_attn.kv_b_proj",
+                "model.layers.45.self_attn.q_a_layernorm",
+                "model.layers.45.shared_head.norm",
+                "model.layers.5.hc_attn_base",
+                "model.layers.5.hc_attn_fn",
+                "model.layers.5.hc_attn_scale",
+                "model.layers.5.hc_ffn_base",
+                "model.layers.5.hc_ffn_fn",
+                "model.layers.5.hc_ffn_scale",
+                "model.layers.5.input_layernorm",
+                "model.layers.5.mlp.gate",
+                "model.layers.5.mlp.gate.e_score_correction_bias",
+                "model.layers.5.post_attention_layernorm",
+                "model.layers.5.self_attn.A_log",
+                "model.layers.5.self_attn.b_proj",
+                "model.layers.5.self_attn.dt_bias",
+                "model.layers.5.self_attn.f_a_proj",
+                "model.layers.5.self_attn.f_b_proj",
+                "model.layers.5.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.5.self_attn.g_a_proj",
+                "model.layers.5.self_attn.g_b_proj",
+                "model.layers.5.self_attn.k_conv1d",
+                "model.layers.5.self_attn.k_proj",
+                "model.layers.5.self_attn.o_norm",
+                "model.layers.5.self_attn.o_proj",
+                "model.layers.5.self_attn.q_conv1d",
+                "model.layers.5.self_attn.q_proj",
+                "model.layers.5.self_attn.qkv_proj",
+                "model.layers.5.self_attn.v_conv1d",
+                "model.layers.5.self_attn.v_proj",
+                "model.layers.6.hc_attn_base",
+                "model.layers.6.hc_attn_fn",
+                "model.layers.6.hc_attn_scale",
+                "model.layers.6.hc_ffn_base",
+                "model.layers.6.hc_ffn_fn",
+                "model.layers.6.hc_ffn_scale",
+                "model.layers.6.input_layernorm",
+                "model.layers.6.mlp.gate",
+                "model.layers.6.mlp.gate.e_score_correction_bias",
+                "model.layers.6.post_attention_layernorm",
+                "model.layers.6.self_attn.A_log",
+                "model.layers.6.self_attn.b_proj",
+                "model.layers.6.self_attn.dt_bias",
+                "model.layers.6.self_attn.f_a_proj",
+                "model.layers.6.self_attn.f_b_proj",
+                "model.layers.6.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.6.self_attn.g_a_proj",
+                "model.layers.6.self_attn.g_b_proj",
+                "model.layers.6.self_attn.k_conv1d",
+                "model.layers.6.self_attn.k_proj",
+                "model.layers.6.self_attn.o_norm",
+                "model.layers.6.self_attn.o_proj",
+                "model.layers.6.self_attn.q_conv1d",
+                "model.layers.6.self_attn.q_proj",
+                "model.layers.6.self_attn.qkv_proj",
+                "model.layers.6.self_attn.v_conv1d",
+                "model.layers.6.self_attn.v_proj",
+                "model.layers.7.hc_attn_base",
+                "model.layers.7.hc_attn_fn",
+                "model.layers.7.hc_attn_scale",
+                "model.layers.7.hc_ffn_base",
+                "model.layers.7.hc_ffn_fn",
+                "model.layers.7.hc_ffn_scale",
+                "model.layers.7.input_layernorm",
+                "model.layers.7.mlp.gate",
+                "model.layers.7.mlp.gate.e_score_correction_bias",
+                "model.layers.7.post_attention_layernorm",
+                "model.layers.7.self_attn.indexer.index_kpool_compress_ape",
+                "model.layers.7.self_attn.indexer.index_kpool_compress_gate",
+                "model.layers.7.self_attn.indexer.k_norm",
+                "model.layers.7.self_attn.indexer.k_norm.bias",
+                "model.layers.7.self_attn.indexer.weights_proj",
+                "model.layers.7.self_attn.indexer.wk",
+                "model.layers.7.self_attn.indexer.wq_b",
+                "model.layers.7.self_attn.kv_a_layernorm",
+                "model.layers.7.self_attn.kv_b_proj",
+                "model.layers.7.self_attn.q_a_layernorm",
+                "model.layers.8.hc_attn_base",
+                "model.layers.8.hc_attn_fn",
+                "model.layers.8.hc_attn_scale",
+                "model.layers.8.hc_ffn_base",
+                "model.layers.8.hc_ffn_fn",
+                "model.layers.8.hc_ffn_scale",
+                "model.layers.8.input_layernorm",
+                "model.layers.8.mlp.gate",
+                "model.layers.8.mlp.gate.e_score_correction_bias",
+                "model.layers.8.post_attention_layernorm",
+                "model.layers.8.self_attn.A_log",
+                "model.layers.8.self_attn.b_proj",
+                "model.layers.8.self_attn.dt_bias",
+                "model.layers.8.self_attn.f_a_proj",
+                "model.layers.8.self_attn.f_b_proj",
+                "model.layers.8.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.8.self_attn.g_a_proj",
+                "model.layers.8.self_attn.g_b_proj",
+                "model.layers.8.self_attn.k_conv1d",
+                "model.layers.8.self_attn.k_proj",
+                "model.layers.8.self_attn.o_norm",
+                "model.layers.8.self_attn.o_proj",
+                "model.layers.8.self_attn.q_conv1d",
+                "model.layers.8.self_attn.q_proj",
+                "model.layers.8.self_attn.qkv_proj",
+                "model.layers.8.self_attn.v_conv1d",
+                "model.layers.8.self_attn.v_proj",
+                "model.layers.9.hc_attn_base",
+                "model.layers.9.hc_attn_fn",
+                "model.layers.9.hc_attn_scale",
+                "model.layers.9.hc_ffn_base",
+                "model.layers.9.hc_ffn_fn",
+                "model.layers.9.hc_ffn_scale",
+                "model.layers.9.input_layernorm",
+                "model.layers.9.mlp.gate",
+                "model.layers.9.mlp.gate.e_score_correction_bias",
+                "model.layers.9.post_attention_layernorm",
+                "model.layers.9.self_attn.A_log",
+                "model.layers.9.self_attn.b_proj",
+                "model.layers.9.self_attn.dt_bias",
+                "model.layers.9.self_attn.f_a_proj",
+                "model.layers.9.self_attn.f_b_proj",
+                "model.layers.9.self_attn.fused_qkvbfg_a_proj",
+                "model.layers.9.self_attn.g_a_proj",
+                "model.layers.9.self_attn.g_b_proj",
+                "model.layers.9.self_attn.k_conv1d",
+                "model.layers.9.self_attn.k_proj",
+                "model.layers.9.self_attn.o_norm",
+                "model.layers.9.self_attn.o_proj",
+                "model.layers.9.self_attn.q_conv1d",
+                "model.layers.9.self_attn.q_proj",
+                "model.layers.9.self_attn.qkv_proj",
+                "model.layers.9.self_attn.v_conv1d",
+                "model.layers.9.self_attn.v_proj",
+                "model.norm",
+                "model.visual",
+                "router",
+                "visual",
+                "visual.blocks.0.attn.k_norm",
+                "visual.blocks.0.attn.proj",
+                "visual.blocks.0.attn.proj.bias",
+                "visual.blocks.0.attn.q_norm",
+                "visual.blocks.0.attn.qkv",
+                "visual.blocks.0.attn.qkv.bias",
+                "visual.blocks.0.mlp.down_proj",
+                "visual.blocks.0.mlp.down_proj.bias",
+                "visual.blocks.0.mlp.gate_proj",
+                "visual.blocks.0.mlp.gate_proj.bias",
+                "visual.blocks.0.mlp.up_proj",
+                "visual.blocks.0.mlp.up_proj.bias",
+                "visual.blocks.0.norm1",
+                "visual.blocks.0.norm2",
+                "visual.blocks.1.attn.k_norm",
+                "visual.blocks.1.attn.proj",
+                "visual.blocks.1.attn.proj.bias",
+                "visual.blocks.1.attn.q_norm",
+                "visual.blocks.1.attn.qkv",
+                "visual.blocks.1.attn.qkv.bias",
+                "visual.blocks.1.mlp.down_proj",
+                "visual.blocks.1.mlp.down_proj.bias",
+                "visual.blocks.1.mlp.gate_proj",
+                "visual.blocks.1.mlp.gate_proj.bias",
+                "visual.blocks.1.mlp.up_proj",
+                "visual.blocks.1.mlp.up_proj.bias",
+                "visual.blocks.1.norm1",
+                "visual.blocks.1.norm2",
+                "visual.blocks.10.attn.k_norm",
+                "visual.blocks.10.attn.proj",
+                "visual.blocks.10.attn.proj.bias",
+                "visual.blocks.10.attn.q_norm",
+                "visual.blocks.10.attn.qkv",
+                "visual.blocks.10.attn.qkv.bias",
+                "visual.blocks.10.mlp.down_proj",
+                "visual.blocks.10.mlp.down_proj.bias",
+                "visual.blocks.10.mlp.gate_proj",
+                "visual.blocks.10.mlp.gate_proj.bias",
+                "visual.blocks.10.mlp.up_proj",
+                "visual.blocks.10.mlp.up_proj.bias",
+                "visual.blocks.10.norm1",
+                "visual.blocks.10.norm2",
+                "visual.blocks.11.attn.k_norm",
+                "visual.blocks.11.attn.proj",
+                "visual.blocks.11.attn.proj.bias",
+                "visual.blocks.11.attn.q_norm",
+                "visual.blocks.11.attn.qkv",
+                "visual.blocks.11.attn.qkv.bias",
+                "visual.blocks.11.mlp.down_proj",
+                "visual.blocks.11.mlp.down_proj.bias",
+                "visual.blocks.11.mlp.gate_proj",
+                "visual.blocks.11.mlp.gate_proj.bias",
+                "visual.blocks.11.mlp.up_proj",
+                "visual.blocks.11.mlp.up_proj.bias",
+                "visual.blocks.11.norm1",
+                "visual.blocks.11.norm2",
+                "visual.blocks.12.attn.k_norm",
+                "visual.blocks.12.attn.proj",
+                "visual.blocks.12.attn.proj.bias",
+                "visual.blocks.12.attn.q_norm",
+                "visual.blocks.12.attn.qkv",
+                "visual.blocks.12.attn.qkv.bias",
+                "visual.blocks.12.mlp.down_proj",
+                "visual.blocks.12.mlp.down_proj.bias",
+                "visual.blocks.12.mlp.gate_proj",
+                "visual.blocks.12.mlp.gate_proj.bias",
+                "visual.blocks.12.mlp.up_proj",
+                "visual.blocks.12.mlp.up_proj.bias",
+                "visual.blocks.12.norm1",
+                "visual.blocks.12.norm2",
+                "visual.blocks.13.attn.k_norm",
+                "visual.blocks.13.attn.proj",
+                "visual.blocks.13.attn.proj.bias",
+                "visual.blocks.13.attn.q_norm",
+                "visual.blocks.13.attn.qkv",
+                "visual.blocks.13.attn.qkv.bias",
+                "visual.blocks.13.mlp.down_proj",
+                "visual.blocks.13.mlp.down_proj.bias",
+                "visual.blocks.13.mlp.gate_proj",
+                "visual.blocks.13.mlp.gate_proj.bias",
+                "visual.blocks.13.mlp.up_proj",
+                "visual.blocks.13.mlp.up_proj.bias",
+                "visual.blocks.13.norm1",
+                "visual.blocks.13.norm2",
+                "visual.blocks.14.attn.k_norm",
+                "visual.blocks.14.attn.proj",
+                "visual.blocks.14.attn.proj.bias",
+                "visual.blocks.14.attn.q_norm",
+                "visual.blocks.14.attn.qkv",
+                "visual.blocks.14.attn.qkv.bias",
+                "visual.blocks.14.mlp.down_proj",
+                "visual.blocks.14.mlp.down_proj.bias",
+                "visual.blocks.14.mlp.gate_proj",
+                "visual.blocks.14.mlp.gate_proj.bias",
+                "visual.blocks.14.mlp.up_proj",
+                "visual.blocks.14.mlp.up_proj.bias",
+                "visual.blocks.14.norm1",
+                "visual.blocks.14.norm2",
+                "visual.blocks.15.attn.k_norm",
+                "visual.blocks.15.attn.proj",
+                "visual.blocks.15.attn.proj.bias",
+                "visual.blocks.15.attn.q_norm",
+                "visual.blocks.15.attn.qkv",
+                "visual.blocks.15.attn.qkv.bias",
+                "visual.blocks.15.mlp.down_proj",
+                "visual.blocks.15.mlp.down_proj.bias",
+                "visual.blocks.15.mlp.gate_proj",
+                "visual.blocks.15.mlp.gate_proj.bias",
+                "visual.blocks.15.mlp.up_proj",
+                "visual.blocks.15.mlp.up_proj.bias",
+                "visual.blocks.15.norm1",
+                "visual.blocks.15.norm2",
+                "visual.blocks.16.attn.k_norm",
+                "visual.blocks.16.attn.proj",
+                "visual.blocks.16.attn.proj.bias",
+                "visual.blocks.16.attn.q_norm",
+                "visual.blocks.16.attn.qkv",
+                "visual.blocks.16.attn.qkv.bias",
+                "visual.blocks.16.mlp.down_proj",
+                "visual.blocks.16.mlp.down_proj.bias",
+                "visual.blocks.16.mlp.gate_proj",
+                "visual.blocks.16.mlp.gate_proj.bias",
+                "visual.blocks.16.mlp.up_proj",
+                "visual.blocks.16.mlp.up_proj.bias",
+                "visual.blocks.16.norm1",
+                "visual.blocks.16.norm2",
+                "visual.blocks.17.attn.k_norm",
+                "visual.blocks.17.attn.proj",
+                "visual.blocks.17.attn.proj.bias",
+                "visual.blocks.17.attn.q_norm",
+                "visual.blocks.17.attn.qkv",
+                "visual.blocks.17.attn.qkv.bias",
+                "visual.blocks.17.mlp.down_proj",
+                "visual.blocks.17.mlp.down_proj.bias",
+                "visual.blocks.17.mlp.gate_proj",
+                "visual.blocks.17.mlp.gate_proj.bias",
+                "visual.blocks.17.mlp.up_proj",
+                "visual.blocks.17.mlp.up_proj.bias",
+                "visual.blocks.17.norm1",
+                "visual.blocks.17.norm2",
+                "visual.blocks.18.attn.k_norm",
+                "visual.blocks.18.attn.proj",
+                "visual.blocks.18.attn.proj.bias",
+                "visual.blocks.18.attn.q_norm",
+                "visual.blocks.18.attn.qkv",
+                "visual.blocks.18.attn.qkv.bias",
+                "visual.blocks.18.mlp.down_proj",
+                "visual.blocks.18.mlp.down_proj.bias",
+                "visual.blocks.18.mlp.gate_proj",
+                "visual.blocks.18.mlp.gate_proj.bias",
+                "visual.blocks.18.mlp.up_proj",
+                "visual.blocks.18.mlp.up_proj.bias",
+                "visual.blocks.18.norm1",
+                "visual.blocks.18.norm2",
+                "visual.blocks.19.attn.k_norm",
+                "visual.blocks.19.attn.proj",
+                "visual.blocks.19.attn.proj.bias",
+                "visual.blocks.19.attn.q_norm",
+                "visual.blocks.19.attn.qkv",
+                "visual.blocks.19.attn.qkv.bias",
+                "visual.blocks.19.mlp.down_proj",
+                "visual.blocks.19.mlp.down_proj.bias",
+                "visual.blocks.19.mlp.gate_proj",
+                "visual.blocks.19.mlp.gate_proj.bias",
+                "visual.blocks.19.mlp.up_proj",
+                "visual.blocks.19.mlp.up_proj.bias",
+                "visual.blocks.19.norm1",
+                "visual.blocks.19.norm2",
+                "visual.blocks.2.attn.k_norm",
+                "visual.blocks.2.attn.proj",
+                "visual.blocks.2.attn.proj.bias",
+                "visual.blocks.2.attn.q_norm",
+                "visual.blocks.2.attn.qkv",
+                "visual.blocks.2.attn.qkv.bias",
+                "visual.blocks.2.mlp.down_proj",
+                "visual.blocks.2.mlp.down_proj.bias",
+                "visual.blocks.2.mlp.gate_proj",
+                "visual.blocks.2.mlp.gate_proj.bias",
+                "visual.blocks.2.mlp.up_proj",
+                "visual.blocks.2.mlp.up_proj.bias",
+                "visual.blocks.2.norm1",
+                "visual.blocks.2.norm2",
+                "visual.blocks.20.attn.k_norm",
+                "visual.blocks.20.attn.proj",
+                "visual.blocks.20.attn.proj.bias",
+                "visual.blocks.20.attn.q_norm",
+                "visual.blocks.20.attn.qkv",
+                "visual.blocks.20.attn.qkv.bias",
+                "visual.blocks.20.mlp.down_proj",
+                "visual.blocks.20.mlp.down_proj.bias",
+                "visual.blocks.20.mlp.gate_proj",
+                "visual.blocks.20.mlp.gate_proj.bias",
+                "visual.blocks.20.mlp.up_proj",
+                "visual.blocks.20.mlp.up_proj.bias",
+                "visual.blocks.20.norm1",
+                "visual.blocks.20.norm2",
+                "visual.blocks.21.attn.k_norm",
+                "visual.blocks.21.attn.proj",
+                "visual.blocks.21.attn.proj.bias",
+                "visual.blocks.21.attn.q_norm",
+                "visual.blocks.21.attn.qkv",
+                "visual.blocks.21.attn.qkv.bias",
+                "visual.blocks.21.mlp.down_proj",
+                "visual.blocks.21.mlp.down_proj.bias",
+                "visual.blocks.21.mlp.gate_proj",
+                "visual.blocks.21.mlp.gate_proj.bias",
+                "visual.blocks.21.mlp.up_proj",
+                "visual.blocks.21.mlp.up_proj.bias",
+                "visual.blocks.21.norm1",
+                "visual.blocks.21.norm2",
+                "visual.blocks.22.attn.k_norm",
+                "visual.blocks.22.attn.proj",
+                "visual.blocks.22.attn.proj.bias",
+                "visual.blocks.22.attn.q_norm",
+                "visual.blocks.22.attn.qkv",
+                "visual.blocks.22.attn.qkv.bias",
+                "visual.blocks.22.mlp.down_proj",
+                "visual.blocks.22.mlp.down_proj.bias",
+                "visual.blocks.22.mlp.gate_proj",
+                "visual.blocks.22.mlp.gate_proj.bias",
+                "visual.blocks.22.mlp.up_proj",
+                "visual.blocks.22.mlp.up_proj.bias",
+                "visual.blocks.22.norm1",
+                "visual.blocks.22.norm2",
+                "visual.blocks.23.attn.k_norm",
+                "visual.blocks.23.attn.proj",
+                "visual.blocks.23.attn.proj.bias",
+                "visual.blocks.23.attn.q_norm",
+                "visual.blocks.23.attn.qkv",
+                "visual.blocks.23.attn.qkv.bias",
+                "visual.blocks.23.mlp.down_proj",
+                "visual.blocks.23.mlp.down_proj.bias",
+                "visual.blocks.23.mlp.gate_proj",
+                "visual.blocks.23.mlp.gate_proj.bias",
+                "visual.blocks.23.mlp.up_proj",
+                "visual.blocks.23.mlp.up_proj.bias",
+                "visual.blocks.23.norm1",
+                "visual.blocks.23.norm2",
+                "visual.blocks.3.attn.k_norm",
+                "visual.blocks.3.attn.proj",
+                "visual.blocks.3.attn.proj.bias",
+                "visual.blocks.3.attn.q_norm",
+                "visual.blocks.3.attn.qkv",
+                "visual.blocks.3.attn.qkv.bias",
+                "visual.blocks.3.mlp.down_proj",
+                "visual.blocks.3.mlp.down_proj.bias",
+                "visual.blocks.3.mlp.gate_proj",
+                "visual.blocks.3.mlp.gate_proj.bias",
+                "visual.blocks.3.mlp.up_proj",
+                "visual.blocks.3.mlp.up_proj.bias",
+                "visual.blocks.3.norm1",
+                "visual.blocks.3.norm2",
+                "visual.blocks.4.attn.k_norm",
+                "visual.blocks.4.attn.proj",
+                "visual.blocks.4.attn.proj.bias",
+                "visual.blocks.4.attn.q_norm",
+                "visual.blocks.4.attn.qkv",
+                "visual.blocks.4.attn.qkv.bias",
+                "visual.blocks.4.mlp.down_proj",
+                "visual.blocks.4.mlp.down_proj.bias",
+                "visual.blocks.4.mlp.gate_proj",
+                "visual.blocks.4.mlp.gate_proj.bias",
+                "visual.blocks.4.mlp.up_proj",
+                "visual.blocks.4.mlp.up_proj.bias",
+                "visual.blocks.4.norm1",
+                "visual.blocks.4.norm2",
+                "visual.blocks.5.attn.k_norm",
+                "visual.blocks.5.attn.proj",
+                "visual.blocks.5.attn.proj.bias",
+                "visual.blocks.5.attn.q_norm",
+                "visual.blocks.5.attn.qkv",
+                "visual.blocks.5.attn.qkv.bias",
+                "visual.blocks.5.mlp.down_proj",
+                "visual.blocks.5.mlp.down_proj.bias",
+                "visual.blocks.5.mlp.gate_proj",
+                "visual.blocks.5.mlp.gate_proj.bias",
+                "visual.blocks.5.mlp.up_proj",
+                "visual.blocks.5.mlp.up_proj.bias",
+                "visual.blocks.5.norm1",
+                "visual.blocks.5.norm2",
+                "visual.blocks.6.attn.k_norm",
+                "visual.blocks.6.attn.proj",
+                "visual.blocks.6.attn.proj.bias",
+                "visual.blocks.6.attn.q_norm",
+                "visual.blocks.6.attn.qkv",
+                "visual.blocks.6.attn.qkv.bias",
+                "visual.blocks.6.mlp.down_proj",
+                "visual.blocks.6.mlp.down_proj.bias",
+                "visual.blocks.6.mlp.gate_proj",
+                "visual.blocks.6.mlp.gate_proj.bias",
+                "visual.blocks.6.mlp.up_proj",
+                "visual.blocks.6.mlp.up_proj.bias",
+                "visual.blocks.6.norm1",
+                "visual.blocks.6.norm2",
+                "visual.blocks.7.attn.k_norm",
+                "visual.blocks.7.attn.proj",
+                "visual.blocks.7.attn.proj.bias",
+                "visual.blocks.7.attn.q_norm",
+                "visual.blocks.7.attn.qkv",
+                "visual.blocks.7.attn.qkv.bias",
+                "visual.blocks.7.mlp.down_proj",
+                "visual.blocks.7.mlp.down_proj.bias",
+                "visual.blocks.7.mlp.gate_proj",
+                "visual.blocks.7.mlp.gate_proj.bias",
+                "visual.blocks.7.mlp.up_proj",
+                "visual.blocks.7.mlp.up_proj.bias",
+                "visual.blocks.7.norm1",
+                "visual.blocks.7.norm2",
+                "visual.blocks.8.attn.k_norm",
+                "visual.blocks.8.attn.proj",
+                "visual.blocks.8.attn.proj.bias",
+                "visual.blocks.8.attn.q_norm",
+                "visual.blocks.8.attn.qkv",
+                "visual.blocks.8.attn.qkv.bias",
+                "visual.blocks.8.mlp.down_proj",
+                "visual.blocks.8.mlp.down_proj.bias",
+                "visual.blocks.8.mlp.gate_proj",
+                "visual.blocks.8.mlp.gate_proj.bias",
+                "visual.blocks.8.mlp.up_proj",
+                "visual.blocks.8.mlp.up_proj.bias",
+                "visual.blocks.8.norm1",
+                "visual.blocks.8.norm2",
+                "visual.blocks.9.attn.k_norm",
+                "visual.blocks.9.attn.proj",
+                "visual.blocks.9.attn.proj.bias",
+                "visual.blocks.9.attn.q_norm",
+                "visual.blocks.9.attn.qkv",
+                "visual.blocks.9.attn.qkv.bias",
+                "visual.blocks.9.mlp.down_proj",
+                "visual.blocks.9.mlp.down_proj.bias",
+                "visual.blocks.9.mlp.gate_proj",
+                "visual.blocks.9.mlp.gate_proj.bias",
+                "visual.blocks.9.mlp.up_proj",
+                "visual.blocks.9.mlp.up_proj.bias",
+                "visual.blocks.9.norm1",
+                "visual.blocks.9.norm2",
+                "visual.downsample",
+                "visual.downsample.bias",
+                "visual.merger.down_proj",
+                "visual.merger.gate_proj",
+                "visual.merger.post_projection_norm",
+                "visual.merger.post_projection_norm.bias",
+                "visual.merger.proj",
+                "visual.merger.up_proj",
+                "visual.patch_embed.proj",
+                "visual.patch_embed.proj.bias",
+                "visual.post_layernorm",
+                "weights_proj"
+            ],
+            "quant_method": "fp8",
+            "weight_block_size": [
+                128,
+                128
+            ]
+        },
+        "vision_bits": 6,
+        "mtp_bits": 4
+    }
+}

generation_config.json

@@ -0,0 +1,12 @@
+{
+  "_from_model_config": true,
+  "eos_token_id": [
+    154820,
+    154827,
+    154829
+  ],
+  "pad_token_id": 154820,
+  "temperature": 1.0,
+  "top_p": 0.95,
+  "transformers_version": "5.16.0"
+}

model-00001-of-00023.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8bb78d5f76c7e2e36c1bd911f59bfb31786f13f48eb3feeaa451d20290095cbe
+size 5660704024

model-00002-of-00023.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4170dff08c925aa36dc28be2c713885947378f87616741e092c6e877616c61a9
+size 7534372728

model-00003-of-00023.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac5d61b63d7f0454104ead2721d5f440aff5901e5c46b057f5636cc2a7384d96
+size 7542828546

model-00004-of-00023.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c9b618ad428185268771c43b781f3c035f6b42a846334ec6d1b4833e67233558
+size 7534372728

model-00005-of-00023.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:16748d78ebad980ebd76971e3ed4e0799056e42c915577b4420f4fa6d2ef4bee
+size 7542835548

model-00006-of-00023.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ea39a5214a6bcf2c03fd76c902cbf2a63b64ae59eaf86bb44ae41c7523feaf7d
+size 7534379718

model-00007-of-00023.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6af72d41509a8b9fbab1a9faffa43c09243f3d45fb9f2e1e675e58b9946dcf25
+size 7542835548

model-00008-of-00023.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:87da44852f3c243be2632e3b97f201b36f8e1eecfe278f9b43cc951252dfa1cb
+size 7534379718

model-00009-of-00023.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b7a9de019bd9492545f9ca2d6a4177dc5061e86223a25c3bbf8aa0342f40c501
+size 7542835548

model-00010-of-00023.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:facb0fa89292c5f100f03d0d790d12bd312fc9c66265292f13e7e43536e9545b
+size 7534379718

model-00011-of-00023.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:155064128a7a1d84e5e76b5bffb0a90a9462450042d80acbcace63ce94bb8769
+size 7542835548

model-00012-of-00023.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f4729ae057bde76acea053e1449f6b98ea108644534d49bc29b8db535aede4ce
+size 7534379718

model-00013-of-00023.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a4c56022293e99263eee276c4a4cee47d7be34c8b2f3322f209609b77075d96f
+size 7542835548

model-00014-of-00023.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:35afef745103c944091b1eac208f6d0602d733dd1f2b1dfb6bcb2949b880b2d6
+size 7534379718

model-00015-of-00023.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6cfd981e5c00786b20736d1e266f96c0b0e79a3d2aaff7ec040aefb66ed436af
+size 7542835548

model-00016-of-00023.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ef5e25b18ebaafaed016366554146f19b110bd3263ecb5eb6565e220589b446a
+size 7534379718

model-00017-of-00023.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:556b1acc875a560090228cb41f699d4ee4a50ef2f3fc05be0e83dbfe25848c5a
+size 7542835548

model-00018-of-00023.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e23f4b82cd8f2352e7db4b9daf420415f114a18e8f81a339bf115694d3cba63
+size 7534379718

model-00019-of-00023.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3636f3e6cc127615bc4c2f08b729842e667cd7722a81d9a48a38069329c8e558
+size 7542835548

model-00020-of-00023.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:988fa5b27f01ee076cd792dfb7db295e5f6ea0e4dd4fd65988d3799272313204
+size 7534379718

model-00021-of-00023.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19d1334c6668b8bea6209ab4c14b4ab05b08ef575aff5d913695d1427f218f20
+size 7542835548

model-00022-of-00023.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f516349d6ca0adcff58b45d4beea4b5dba555b89a89fd69f3c375f53cbec0429
+size 8565726891

model-00023-of-00023.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7367136600dc34d1228e1f1c8751a1560ed11abc687eaa35a0f8b750ab7737c0
+size 69269741

model.safetensors.index.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dd47a4a9e5d47971a3cb7df66c86b83ea23b25b41a776c3b0f031c95f0c13d1b
+size 16343098

processor_config.json

@@ -0,0 +1,44 @@
+{
+  "image_processor": {
+    "do_rescale": true,
+    "patch_expand_factor": 1,
+    "merge_size": 2,
+    "image_mean": [
+      0.48145466,
+      0.4578275,
+      0.40821073
+    ],
+    "image_std": [
+      0.26862954,
+      0.26130258,
+      0.27577711
+    ],
+    "temporal_patch_size": 2,
+    "patch_size": 14,
+    "min_image_tokens": 16,
+    "max_image_tokens": 8000,
+    "image_processor_type": "Glm5NextImageProcessor"
+  },
+  "video_processor": {
+    "do_rescale": true,
+    "video_processor_type": "Glm5NextVideoProcessor",
+    "patch_expand_factor": 1,
+    "merge_size": 2,
+    "image_mean": [
+      0.48145466,
+      0.4578275,
+      0.40821073
+    ],
+    "image_std": [
+      0.26862954,
+      0.26130258,
+      0.27577711
+    ],
+    "temporal_patch_size": 2,
+    "patch_size": 14,
+    "min_image_tokens": 16,
+    "max_image_tokens": 240000,
+    "fps": 2
+  },
+  "processor_class": "Glm5NextProcessor"
+}

quantization_config.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a5a41e535ffa1b0b386b71d4fd2fe83b4faa52b7b1f53e2e294bcb3ee941352
+size 47905719

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19e773648cb4e65de8660ea6365e10acca112d42a854923df93db4a6f333a82d
+size 20217442

tokenizer_config.json

@@ -0,0 +1,33 @@
+{
+  "backend": "tokenizers",
+  "clean_up_tokenization_spaces": false,
+  "do_lower_case": false,
+  "eos_token": "<|endoftext|>",
+  "extra_special_tokens": [
+    "<|endoftext|>",
+    "[MASK]",
+    "[gMASK]",
+    "[sMASK]",
+    "<sop>",
+    "<eop>",
+    "<|system|>",
+    "<|user|>",
+    "<|assistant|>",
+    "<|observation|>",
+    "<|begin_of_image|>",
+    "<|end_of_image|>",
+    "<|begin_of_video|>",
+    "<|end_of_video|>",
+    "<|begin_of_audio|>",
+    "<|end_of_audio|>",
+    "<|begin_of_transcription|>",
+    "<|end_of_transcription|>"
+  ],
+  "is_local": true,
+  "model_max_length": 1048576,
+  "model_specific_special_tokens": {},
+  "pad_token": "<|endoftext|>",
+  "padding_side": "left",
+  "remove_space": false,
+  "tokenizer_class": "TokenizersBackend"
+}