Weights release

.gitattributes

@@ -0,0 +1,36 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz 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

README.md

@@ -0,0 +1,77 @@
+---
+license: cc-by-nc-sa-4.0
+base_model: Qwen/Qwen3.5-9B
+library_name: peft
+pipeline_tag: video-text-to-text
+language:
+- en
+tags:
+- surgical-video
+- video-question-answering
+- long-video-understanding
+- medical
+- lora
+- qwen3.5
+- miccai-2026
+datasets:
+- orena-dkfz/heico-focus-vqa
+- orena-dkfz/lapchole-focus-vqa
+---
+
+# SurgScope
+
+Our solution to the PROCEDURE track of the [ORena SAVE FOCUS Challenge](https://orena-focus-challenge.org/), MICCAI 2026.
+
+LoRA adapter for [Qwen3.5-9B](https://huggingface.co/Qwen/Qwen3.5-9B) that answers questions
+about foreign objects over entire surgical procedures of up to five hours.
+**Code:** [github.com/wxyi057/orena-SurgScope](https://github.com/wxyi057/orena-SurgScope) ·
+**Data:** [HeiCo-FOCUS-VQA](https://huggingface.co/datasets/orena-dkfz/heico-focus-vqa) ·
+[LapChole-FOCUS-VQA](https://huggingface.co/datasets/orena-dkfz/lapchole-focus-vqa)
+
+## Method
+
+19 rules on the question text pick the time window. Every question gets ~46k visual tokens:
+timestamp and counting questions spend them on 1536 frames, all others on 1152 sharper frames.
+A timestamp answer is re-asked on ±10 min and ±2 min windows around it. The adapter averages
+three fine-tunes trained with different frame-sampling regimes.
+
+## Results
+
+| Pre-evaluation | In-distribution | Out-of-distribution |
+|:-:|:-:|:-:|
+| **0.6592** | 0.7195 | 0.5990 |
+
+## Usage
+
+```bash
+git clone https://github.com/wxyi057/orena-SurgScope && cd orena-SurgScope
+pip install -e .
+hf auth login                  # after access to HeiCo-FOCUS-VQA is approved
+bash scripts/make_examples.sh
+surgscope infer --input examples/test --output answer.json
+```
+
+The `surgscope` package reproduces the training-time inputs (windows, frame counts, prompt) and
+merges the adapter into the base model at load time.
+
+## Training
+
+PROCEDURE training split of HeiCo-FOCUS-VQA and LapChole-FOCUS-VQA (6,873 questions). LoRA
+r 64 / α 128 on all linear layers, 15 epochs, ms-swift 4.3.2, three frame-sampling regimes
+(768; 1536 / 1152; 2880 / 1512 frames). Full recipe: `surgscope_recipe.json`.
+
+## License
+
+CC BY-NC-SA 4.0 (base model: Apache 2.0).
+
+## Citation
+
+```bibtex
+@misc{surgscope2026,
+  title  = {SurgScope: Question-Conditioned Windowing for Hour-Long Surgical Video Question Answering},
+  author = {Yi, Weixi and Zhang, Hanyuan and He, Runlong},
+  year   = {2026},
+  note   = {Solution to the PROCEDURE track, ORena SAVE FOCUS Challenge, MICCAI 2026},
+  url    = {https://github.com/wxyi057/orena-SurgScope}
+}
+```

adapter_config.json

@@ -0,0 +1,40 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-9B",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": [],
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": "^(model\\.language_model(?=\\.).*\\.(in_proj_z|o_proj|k_proj|in_proj_qkv|out_proj|gate_proj|up_proj|q_proj|in_proj_b|v_proj|in_proj_a|down_proj)|(?!(model.visual.merger))model\\.visual(?=\\.).*\\.(linear_fc1|linear_fc2|qkv|attn.proj)|model\\.visual\\.merger(?=\\.).*\\.(linear_fc1|linear_fc2))$",
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3ee1bd381a6af9a1d5863ea51ce3c2fc2479e860dbf75c50719870fecca4acfb
+size 820346568

surgscope_recipe.json

@@ -0,0 +1,21 @@
+{
+  "base_model": "Qwen/Qwen3.5-9B",
+  "adapter": {"type": "LoRA", "rank": 64, "alpha": 128, "dropout": 0.05,
+              "target_modules": "all linear layers of vision encoder, patch merger and language model"},
+  "weights": "equal-weight fp32 mean of LoRA A and B of three adapters trained under three frame-sampling regimes (same base, seed and LoRA shape); merged into the base model in bf16 at load time",
+  "regimes": {
+    "r768":        {"frames": "768 for every question", "tokens_per_frame_pair": "<= 128",
+                    "lr": 1.4e-4, "global_batch": 64, "max_length": 57344, "checkpoint_epoch": 4},
+    "dense-mix":   {"frames": "1536 time / counting / aggregation, 1152 otherwise", "tokens_per_frame_pair": "60 / 84",
+                    "lr": 1e-4, "global_batch": 32, "max_length": 57344, "checkpoint_epoch": 14},
+    "dense-mix-2": {"frames": "2880 time / counting / aggregation, 1512 otherwise", "tokens_per_frame_pair": "32 / 64",
+                    "lr": 1e-4, "global_batch": 32, "max_length": 65536, "checkpoint_epoch": 6}
+  },
+  "optimisation": {"optimizer": "AdamW", "weight_decay": 0.1, "adam_beta2": 0.95, "schedule": "cosine",
+                   "warmup_ratio": 0.03, "epochs": 15, "precision": "bf16", "seed": 42, "framework": "ms-swift 4.3.2"},
+  "training_data": "official PROCEDURE train split of HeiCo-FOCUS-VQA and LapChole-FOCUS-VQA (6,873 question-answer pairs), no external data",
+  "inference": {"windows": "19 rules on the question text", "regime": "dense-mix",
+                "refinement": "timestamp answers re-asked on +-10 min and +-2 min windows",
+                "env": {"VIDEO_MAX_TOKEN_NUM": "128", "VIDEO_MIN_TOKEN_NUM": "32", "FORCE_QWENVL_VIDEO_READER": "torchvision"},
+                "decoding": "greedy, max 64 new tokens, thinking disabled"}
+}