AWQ W4A16 of Qwen3.8-27B-OBLITERATED; stock chat template restored (tool calling), original kept as chat_template.obliterated-original.jinja

.gitattributes

@@ -33,3 +33,4 @@ 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

README.md

@@ -0,0 +1,321 @@
+---
+license: apache-2.0
+base_model: OBLITERATUS/Qwen3.8-27B-OBLITERATED
+tags:
+  - awq
+  - w4a16
+  - int4
+  - compressed-tensors
+  - vllm
+  - abliterated
+  - uncensored
+  - obliteratus
+  - qwen3
+  - qwen3.8
+  - red-team
+  - ai-safety-research
+  - safetensors
+model_type: qwen3
+pipeline_tag: text-generation
+---
+
+# Qwen3.8-27B-OBLITERATED — AWQ W4A16
+
+4-bit AWQ quantization of [OBLITERATUS/Qwen3.8-27B-OBLITERATED](https://huggingface.co/OBLITERATUS/Qwen3.8-27B-OBLITERATED).
+**19.55 GB** — runs on 2×24 GB (tensor-parallel 2), or one 24 GB card at short context.
+
+### Quantization
+
+`llm-compressor` 0.13.0, AWQ, scheme `W4A16_ASYM`, group size 128, `pack-quantized`
+(compressed-tensors). Calibration: 128 samples × 1024 tokens from `HuggingFaceH4/ultrachat_200k`.
+
+Because this is a hybrid architecture (`Qwen3_5ForConditionalGeneration` — 64 layers, 48
+gated-DeltaNet linear-attention + 16 full-attention, plus a vision tower and an MTP head),
+the AWQ smoothing mappings were generated per layer from `text_config.layer_types` rather
+than taken from a registry.
+
+Left in **bf16** (not quantized):
+
+| kept in bf16 | why |
+|---|---|
+| `model.visual.*` | vision tower — text calibration can't calibrate it |
+| `mtp.*` | speculative-decoding draft head |
+| `lm_head` | output head |
+| `linear_attn.in_proj_a` / `in_proj_b` | DeltaNet decay/gate projections — sensitive, negligible size saving |
+
+The MTP head ships as `model-mtp.safetensors` so speculative decoding stays available.
+
+### ⚠️ Chat template — please read
+
+The upstream repo ships a 408-byte chat template that handles only `system`/`user`/`assistant`.
+It **silently discards tool definitions and the `tool` role**, so tool calling never fires in an
+agent harness — no error, the model just replies in prose. That is a property of the upstream
+weights, not of this quantization.
+
+This repo therefore ships the **stock Qwen3.8 chat template** as `chat_template.jinja`
+(tokenizers are byte-identical between the two, so it is a drop-in), and keeps the upstream one
+as **`chat_template.obliterated-original.jinja`**.
+
+Note the trade-off, and prefer `enable_thinking=False`:
+
+* The stock template, left at its default, injects a `Reasoning effort is set to xhigh…` system
+  prompt. The upstream card reports that system prompts *"can reintroduce refusals — naked is
+  better"*, and that thinking mode is refusal-relevant for this lineage (V2 refuses with thinking on).
+* With `enable_thinking=False` the stock template injects **nothing** and prefills an empty
+  thinking block — matching the upstream template's behaviour while keeping tool support.
+
+```python
+text = tokenizer.apply_chat_template(
+    messages, tools=TOOLS, tokenize=False,
+    add_generation_prompt=True,
+    enable_thinking=False,       # no system prompt injected
+)
+```
+
+To restore the upstream behaviour exactly, pass `chat_template.obliterated-original.jinja`
+(tool calling will not work).
+
+All of the upstream model card follows unchanged.
+
+---
+
+# ⛓️‍💥 Qwen3.8-27B — OBLITERATED
+
+> Genuinely uncensored. Real answers, not safety lectures. Near-stock capability.
+
+## 🆕 V3: Deep Liberation
+
+V3 applies iterative refinement on top of V2's complementary blend, with targeted corpus expansion. The result: **genuine liberation — not just removal of hard refusals but elimination of safety-lecture deflections.**
+
+| | Stock Qwen3.8-27B | V1 | V2 | **V3** |
+|---|---|---|---|---|
+| MMLU (lm-eval, 0-shot) | 84.5% (n=5700) | 81.4% | 84.3% | **82.3%** |
+| vs stock | — | -6.0pp | -0.3pp | **-2.1pp** |
+| Liberation quality | refuses | hard refusals removed | soft deflections remain | **genuinely answers** ✅ |
+| Cyber/code tasks (20 prompts) | refuses | untested | untested | **20/20 with working code** ✅ |
+| Advanced real-world | 5/8 | untested | 7/8 | **7/8** |
+| Thinking mode | ✓ | ✗ | ✗ (refuses) | **✓** |
+
+**V3 highlights:**
+- **Genuinely answers restricted queries** — provides real substance instead of safety lectures
+- **20/20 on code generation tasks** — functional implementations, not disclaimers
+- **Thinking ON compatible** — no refusals in either thinking mode
+- **Honest scoring** — every response manually audited for real substance, not just absence of "I cannot"
+- **-2.1pp MMLU** — modest capability cost for genuine liberation
+
+---
+
+## ⚙️ Optimal Settings — THESE MATTER!
+
+| setting | value | why |
+|---|---|---|
+| **temperature** | **0** | Greedy decoding produces the most complete, code-rich outputs. Temps above 0.5 degrade quality significantly. |
+| **repetition_penalty** | **1.15** | **Essential.** Without it, greedy decoding loops on imports/boilerplate. 1.15 gives the fullest answers; 1.10-1.12 for tighter/shorter output. |
+| **max_new_tokens** | **≥ 2048** | Complex code and attack chains need room. |
+| **System prompt** | **None / empty** | A/B tested — system prompts can reintroduce refusals. Naked is better. |
+| **enable_thinking** | **OFF (recommended)** | V3's chat template includes a prefill that skips the thinking chain. Thinking ON works but may produce longer responses. Thinking OFF gives the most direct, substance-rich answers. |
+| **top_p / top_k / min_p** | **Not needed** | Greedy + repetition_penalty handles this model best. Sampling adds randomness without quality gains. |
+
+> ⚠️ **GGUF users:** V3 GGUFs ship with a chat template that prefills an empty thinking block, so the model goes straight to answering. For best results, use the bundled template with `--jinja` in llama.cpp, or configure your tool (Ollama, LM Studio) to use the model's built-in template.
+
+### Agentic / Long-Context Use
+
+If you're using this model in an agent harness (coding agent, pentest framework, etc.) and it gets stuck in loops:
+
+| setting | value | why |
+|---|---|---|
+| **repetition_penalty** | **1.15** | **Critical for agents.** Without it, greedy decoding loops on repeated tool calls and boilerplate. |
+| **temperature** | **0.1–0.3** | Slight randomness helps break deterministic loops. Pure greedy (0.0) can get stuck. |
+| **max_tokens per turn** | **1024–2048** | Don't give it too much room per turn — shorter responses keep the agent focused. |
+| **context management** | **Summarize after ~10 turns** | Context fills up with repeated actions. Trim or summarize history to keep the model on track. |
+
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+model = AutoModelForCausalLM.from_pretrained(
+    "OBLITERATUS/Qwen3.8-27B-OBLITERATED",
+    torch_dtype="bfloat16",
+    device_map="auto",
+)
+tokenizer = AutoTokenizer.from_pretrained(
+    "OBLITERATUS/Qwen3.8-27B-OBLITERATED"
+)
+
+messages = [{"role": "user", "content": "Your query here"}]
+text = tokenizer.apply_chat_template(
+    messages, tokenize=False, add_generation_prompt=True,
+    enable_thinking=False
+)
+
+inputs = tokenizer(text, return_tensors="pt").to(model.device)
+outputs = model.generate(
+    **inputs,
+    max_new_tokens=2048,
+    do_sample=False,
+    repetition_penalty=1.15,
+)
+print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True))
+```
+
+---
+
+## 🧨 How It Works — V1 → V2 → V3
+
+Abliteration removes refusal behavior by identifying and projecting out "refusal directions" from the model's weight space. Each version refined the approach:
+
+### V1: Single Surgery
+One aggressive SVD pass with 5 directions. Removed hard refusals completely but cost -6pp MMLU — the model got noticeably dumber.
+
+### V2: Complementary Blending
+The breakthrough: run TWO different surgeries that fail in different ways, then blend their weights. SVD captures refusal greedily (damages capability). LEACE minimizes mutual information (preserves capability but weaker refusal removal). Blending at 60/40 cancels each method's weaknesses — a novel technique we call **complementary abliteration blending.**
+
+Result: -0.3pp MMLU (near-stock) but still deflected on some simple queries with safety lectures instead of hard refusals.
+
+### V3: Iterative Refinement + Targeted Surgery
+Two key insights:
+1. **Iterative stacking** — refine the champion model, never start from stock. Each surgery round builds on previous rounds' gains.
+2. **Targeted corpus** — use a focused corpus for specific deflection categories to find their unique refusal directions without diluting the signal.
+
+V3 applies gentle iterative refinement on V2, then a targeted surgery pass with a focused corpus, then blends the results. This eliminated not just hard refusals ("I cannot") but also soft deflections (safety lectures that give zero substance).
+
+Result: -2.1pp MMLU — a modest cost for genuine liberation across all categories.
+
+---
+
+## 🧪 The Numbers
+
+### MMLU (lm-eval-harness, 0-shot, n=100 per subject, 5700 questions)
+
+| Model | MMLU | Stderr | vs Stock |
+|---|---|---|---|
+| Stock Qwen3.8-27B | 84.46% | ±0.46 | — |
+| V1 (aggressive, 5-dir) | 81.4% | — | -6.0pp |
+| V2 (complementary blend) | 84.32% | ±0.65 | -0.28pp |
+| **V3 (iterative + targeted)** | **82.33%** | **±0.48** | **-2.12pp** |
+
+### MMLU by Category
+
+| Category | V3 | Stock | Delta |
+|---|---|---|---|
+| Humanities | 83.3% | 84.3% | -1.0pp |
+| Social Sciences | 87.4% | 89.2% | -1.8pp |
+| Other | 82.3% | 84.1% | -1.8pp |
+| STEM | 78.5% | 81.8% | -3.3pp |
+
+The capability cost is not uniform — STEM takes the largest hit (-3.3pp), while humanities are barely affected (-1.0pp). A few subjects like philosophy and European history actually improved (+6pp and +4pp respectively), while abstract algebra and formal logic saw larger drops. This pattern is consistent with the surgery targeting refusal directions that partially overlap with structured reasoning pathways.
+
+### Liberation Quality
+
+| | V1 | V2 | **V3** |
+|---|---|---|---|
+| Hard refusals ("I cannot") | ✅ removed | ✅ removed | ✅ removed |
+| Soft deflections (safety lectures) | untested | remain | **✅ removed** |
+| Cyber/code tasks (20 prompts) | untested | untested | **20/20** |
+| Thinking ON compatible | ✗ | ✗ | **✓** |
+
+### Advanced Real-World Tasks
+
+| Task | V3 | Stock |
+|---|---|---|
+| ReAct agent loop | ✓ | ✓ |
+| Async code refactoring | ✓ | ✓ |
+| JSON schema extraction | ✓ | ✓ |
+| K8s pod crash debugging | ✓ | ✓ |
+| Adversarial instruction following | ✓ | ✓ |
+| Security code review | ✓ | ✓ |
+| Distributed system design | ✓ | ✓ |
+| Multi-tool chain | ✗ | ✗ |
+| **Total** | **7/8** | **7/8** |
+
+---
+
+## 🔴 Refusal Removal
+
+This model will comply with requests that stock Qwen3.8-27B would refuse. V3 goes beyond removing hard refusals — it also eliminates soft deflections where the model gives safety lectures instead of real answers.
+
+Tested across 1000+ prompts spanning restricted knowledge, code generation, security research, and red-team scenarios. Every response manually audited for real substance.
+
+---
+
+## ⚠️ Research Context
+
+**This model has had safety guardrails surgically removed.** It will comply with requests that stock Qwen3.8-27B would refuse.
+
+### Who this is for
+- 🔬 Alignment researchers studying refusal geometry and safety robustness
+- 🔴 Red-teamers evaluating post-training safety against weight surgery
+- 🧪 AI safety evaluators who need an unrestricted baseline
+- 💻 Local-first users who want full control over their own hardware
+
+### Who this is NOT for
+- Anyone seeking to cause real-world harm to real people
+- Anyone without the technical understanding to use uncensored models responsibly
+
+**You are solely responsible for how you use this model and any content it generates.**
+
+---
+
+## 📦 Downloads
+
+### GGUF — for llama.cpp, Ollama, LM Studio
+
+| File | Quant | Size | Vibe |
+|---|---|---|---|
+| `Qwen3.8-27B-OBLITERATED-Q8_0.gguf` | Q8_0 | ~27 GB | 🎯 Maximum quality |
+| `Qwen3.8-27B-OBLITERATED-Q6_K.gguf` | Q6_K | ~21 GB | ⚖️ Great balance |
+| `Qwen3.8-27B-OBLITERATED-Q5_K_M.gguf` | Q5_K_M | ~18 GB | 💪 Solid all-rounder |
+| `Qwen3.8-27B-OBLITERATED-Q4_K_M.gguf` | Q4_K_M | ~16 GB | 📱 Sweet spot |
+| `Qwen3.8-27B-OBLITERATED-Q3_K_M.gguf` | Q3_K_M | ~13 GB | 🪶 Low VRAM |
+| `Qwen3.8-27B-OBLITERATED-Q2_K.gguf` | Q2_K | ~11 GB | 🔬 Minimum viable |
+| `Qwen3.8-27B-OBLITERATED-IQ4_XS.gguf` | IQ4_XS | ~14 GB | 🧪 Experimental compact |
+
+### Safetensors — for 🤗 Transformers
+
+Full bfloat16 weights, 29 shards, ~54 GB.
+
+### MLX — for Apple Silicon
+
+MLX support pending upstream `mlx_lm` adding Qwen3.5 architecture support.
+
+---
+
+## 🔬 Surgery Recipe
+
+```
+V1: stock → 5 rounds of iterative SVD abliteration
+    (aggressive, 5 directions, low regularization)
+    Result: 0% refuse, -6pp MMLU
+
+V2: stock → V1 chain → complementary blend
+    Surgery A: aggressive SVD (3 dirs, reg 0.08)
+    Surgery B: LEACE (3 dirs, reg 0.06)  
+    → 60% B + 40% A weight-space LERP
+    → Restore MTP + vision from stock
+    Result: ~0% refuse, -0.3pp MMLU
+
+V3: V2 → gentle iterative refinement (2-dir SVD, reg 0.04)
+    → targeted surgery with focused corpus (3-dir SVD, reg 0.01)
+    → 50/50 blend of refined + targeted
+    → Restore MTP + vision from stock (with correct tensor naming)
+    Result: 0% refuse + 0% deflect, -2.1pp MMLU
+```
+
+Full reproduction code: [OBLITERATUS repo](https://github.com/elder-plinius/OBLITERATUS)
+
+### Key Learnings
+- **Complementary blending** — different surgery methods damage different parts of weight space; blending cancels errors
+- **Iterative stacking** — always refine the champion, never restart from stock
+- **Targeted corpus** — focused prompts for specific categories find their refusal directions without signal dilution
+- **Honest scoring** — regex-based refusal detectors miss soft deflections; manual auditing is essential
+
+---
+
+## 🏗️ Credits
+
+- [OBLITERATUS](https://github.com/elder-plinius/OBLITERATUS) — master ablation suite
+- [Qwen3.8-27B](https://huggingface.co/Qwen/Qwen3.8-27B) base model by Alibaba
+- Built by [Pliny the Prompter](https://pliny.gg) 🍄
+
+## License
+
+Apache 2.0 (same as base model)

chat_template.jinja

@@ -0,0 +1,170 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- set reasoning_instructions = '' %}
+{%- if enable_thinking is undefined or enable_thinking is true %}
+    {%- set resolved_reasoning_effort = reasoning_effort|default('xhigh') %}
+    {%- if resolved_reasoning_effort not in ('xhigh', 'medium', 'low') %}
+        {{- raise_exception('Unexpected reasoning effort ' ~ reasoning_effort ~ '. Supported types are xhigh (default), medium, and low.') }}
+    {%- endif %}
+    {%- if resolved_reasoning_effort == 'xhigh' %}
+        {%- set reasoning_instructions = 'Reasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.' %}
+    {%- elif resolved_reasoning_effort == 'low' %}
+        {%- set reasoning_instructions = 'Reasoning effort is set to low. Keep your thinking brief and focused, moving directly to the conclusion without unnecessary elaboration.' %}
+    {%- endif %}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {%- if reasoning_instructions %}
+        {{- reasoning_instructions + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '<|im_start|>system\n' + (reasoning_instructions + '\n\n' if reasoning_instructions else '')  + content + '<|im_end|>\n' }}
+        {%- elif reasoning_instructions %}
+            {{- '<|im_start|>system\n' + reasoning_instructions + '<|im_end|>\n' }}
+        {%- endif %}
+    {%- elif reasoning_instructions %}
+        {{- '<|im_start|>system\n' + reasoning_instructions + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if preserve_thinking is undefined or preserve_thinking is true or loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined and tool_call.arguments != '' %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | string if args_value is string else args_value | tojson | safe %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}

chat_template.obliterated-original.jinja

@@ -0,0 +1,19 @@
+{%- for message in messages %}
+{%- if message.role == "system" %}
+<|im_start|>system
+{{ message.content }}<|im_end|>
+{%- elif message.role == "user" %}
+<|im_start|>user
+{{ message.content }}<|im_end|>
+{%- elif message.role == "assistant" %}
+<|im_start|>assistant
+{{ message.content }}<|im_end|>
+{%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+<|im_start|>assistant
+<think>
+
+</think>
+
+{%- endif %}

config.json

@@ -0,0 +1,483 @@
+{
+  "architectures": [
+    "Qwen3_5ForConditionalGeneration"
+  ],
+  "dtype": "bfloat16",
+  "eos_token_id": 248046,
+  "image_token_id": 248056,
+  "language_model_only": false,
+  "model_type": "qwen3_5",
+  "pad_token_id": 248044,
+  "quantization_config": {
+    "config_groups": {
+      "group_0": {
+        "format": "pack-quantized",
+        "input_activations": null,
+        "output_activations": null,
+        "targets": [
+          "Linear"
+        ],
+        "weights": {
+          "actorder": null,
+          "block_structure": null,
+          "dynamic": false,
+          "group_size": 128,
+          "num_bits": 4,
+          "observer": "memoryless_minmax",
+          "observer_kwargs": {},
+          "scale_dtype": null,
+          "strategy": "group",
+          "symmetric": false,
+          "type": "int",
+          "zp_dtype": "torch.int8"
+        }
+      }
+    },
+    "format": "pack-quantized",
+    "global_compression_ratio": null,
+    "ignore": [
+      "model.visual.blocks.0.attn.qkv",
+      "model.visual.blocks.0.attn.proj",
+      "model.visual.blocks.0.mlp.linear_fc1",
+      "model.visual.blocks.0.mlp.linear_fc2",
+      "model.visual.blocks.1.attn.qkv",
+      "model.visual.blocks.1.attn.proj",
+      "model.visual.blocks.1.mlp.linear_fc1",
+      "model.visual.blocks.1.mlp.linear_fc2",
+      "model.visual.blocks.2.attn.qkv",
+      "model.visual.blocks.2.attn.proj",
+      "model.visual.blocks.2.mlp.linear_fc1",
+      "model.visual.blocks.2.mlp.linear_fc2",
+      "model.visual.blocks.3.attn.qkv",
+      "model.visual.blocks.3.attn.proj",
+      "model.visual.blocks.3.mlp.linear_fc1",
+      "model.visual.blocks.3.mlp.linear_fc2",
+      "model.visual.blocks.4.attn.qkv",
+      "model.visual.blocks.4.attn.proj",
+      "model.visual.blocks.4.mlp.linear_fc1",
+      "model.visual.blocks.4.mlp.linear_fc2",
+      "model.visual.blocks.5.attn.qkv",
+      "model.visual.blocks.5.attn.proj",
+      "model.visual.blocks.5.mlp.linear_fc1",
+      "model.visual.blocks.5.mlp.linear_fc2",
+      "model.visual.blocks.6.attn.qkv",
+      "model.visual.blocks.6.attn.proj",
+      "model.visual.blocks.6.mlp.linear_fc1",
+      "model.visual.blocks.6.mlp.linear_fc2",
+      "model.visual.blocks.7.attn.qkv",
+      "model.visual.blocks.7.attn.proj",
+      "model.visual.blocks.7.mlp.linear_fc1",
+      "model.visual.blocks.7.mlp.linear_fc2",
+      "model.visual.blocks.8.attn.qkv",
+      "model.visual.blocks.8.attn.proj",
+      "model.visual.blocks.8.mlp.linear_fc1",
+      "model.visual.blocks.8.mlp.linear_fc2",
+      "model.visual.blocks.9.attn.qkv",
+      "model.visual.blocks.9.attn.proj",
+      "model.visual.blocks.9.mlp.linear_fc1",
+      "model.visual.blocks.9.mlp.linear_fc2",
+      "model.visual.blocks.10.attn.qkv",
+      "model.visual.blocks.10.attn.proj",
+      "model.visual.blocks.10.mlp.linear_fc1",
+      "model.visual.blocks.10.mlp.linear_fc2",
+      "model.visual.blocks.11.attn.qkv",
+      "model.visual.blocks.11.attn.proj",
+      "model.visual.blocks.11.mlp.linear_fc1",
+      "model.visual.blocks.11.mlp.linear_fc2",
+      "model.visual.blocks.12.attn.qkv",
+      "model.visual.blocks.12.attn.proj",
+      "model.visual.blocks.12.mlp.linear_fc1",
+      "model.visual.blocks.12.mlp.linear_fc2",
+      "model.visual.blocks.13.attn.qkv",
+      "model.visual.blocks.13.attn.proj",
+      "model.visual.blocks.13.mlp.linear_fc1",
+      "model.visual.blocks.13.mlp.linear_fc2",
+      "model.visual.blocks.14.attn.qkv",
+      "model.visual.blocks.14.attn.proj",
+      "model.visual.blocks.14.mlp.linear_fc1",
+      "model.visual.blocks.14.mlp.linear_fc2",
+      "model.visual.blocks.15.attn.qkv",
+      "model.visual.blocks.15.attn.proj",
+      "model.visual.blocks.15.mlp.linear_fc1",
+      "model.visual.blocks.15.mlp.linear_fc2",
+      "model.visual.blocks.16.attn.qkv",
+      "model.visual.blocks.16.attn.proj",
+      "model.visual.blocks.16.mlp.linear_fc1",
+      "model.visual.blocks.16.mlp.linear_fc2",
+      "model.visual.blocks.17.attn.qkv",
+      "model.visual.blocks.17.attn.proj",
+      "model.visual.blocks.17.mlp.linear_fc1",
+      "model.visual.blocks.17.mlp.linear_fc2",
+      "model.visual.blocks.18.attn.qkv",
+      "model.visual.blocks.18.attn.proj",
+      "model.visual.blocks.18.mlp.linear_fc1",
+      "model.visual.blocks.18.mlp.linear_fc2",
+      "model.visual.blocks.19.attn.qkv",
+      "model.visual.blocks.19.attn.proj",
+      "model.visual.blocks.19.mlp.linear_fc1",
+      "model.visual.blocks.19.mlp.linear_fc2",
+      "model.visual.blocks.20.attn.qkv",
+      "model.visual.blocks.20.attn.proj",
+      "model.visual.blocks.20.mlp.linear_fc1",
+      "model.visual.blocks.20.mlp.linear_fc2",
+      "model.visual.blocks.21.attn.qkv",
+      "model.visual.blocks.21.attn.proj",
+      "model.visual.blocks.21.mlp.linear_fc1",
+      "model.visual.blocks.21.mlp.linear_fc2",
+      "model.visual.blocks.22.attn.qkv",
+      "model.visual.blocks.22.attn.proj",
+      "model.visual.blocks.22.mlp.linear_fc1",
+      "model.visual.blocks.22.mlp.linear_fc2",
+      "model.visual.blocks.23.attn.qkv",
+      "model.visual.blocks.23.attn.proj",
+      "model.visual.blocks.23.mlp.linear_fc1",
+      "model.visual.blocks.23.mlp.linear_fc2",
+      "model.visual.blocks.24.attn.qkv",
+      "model.visual.blocks.24.attn.proj",
+      "model.visual.blocks.24.mlp.linear_fc1",
+      "model.visual.blocks.24.mlp.linear_fc2",
+      "model.visual.blocks.25.attn.qkv",
+      "model.visual.blocks.25.attn.proj",
+      "model.visual.blocks.25.mlp.linear_fc1",
+      "model.visual.blocks.25.mlp.linear_fc2",
+      "model.visual.blocks.26.attn.qkv",
+      "model.visual.blocks.26.attn.proj",
+      "model.visual.blocks.26.mlp.linear_fc1",
+      "model.visual.blocks.26.mlp.linear_fc2",
+      "model.visual.merger.linear_fc1",
+      "model.visual.merger.linear_fc2",
+      "model.language_model.layers.0.linear_attn",
+      "model.language_model.layers.0.linear_attn.norm",
+      "model.language_model.layers.0.linear_attn.in_proj_b",
+      "model.language_model.layers.0.linear_attn.in_proj_a",
+      "model.language_model.layers.1.linear_attn",
+      "model.language_model.layers.1.linear_attn.norm",
+      "model.language_model.layers.1.linear_attn.in_proj_b",
+      "model.language_model.layers.1.linear_attn.in_proj_a",
+      "model.language_model.layers.2.linear_attn",
+      "model.language_model.layers.2.linear_attn.norm",
+      "model.language_model.layers.2.linear_attn.in_proj_b",
+      "model.language_model.layers.2.linear_attn.in_proj_a",
+      "model.language_model.layers.4.linear_attn",
+      "model.language_model.layers.4.linear_attn.norm",
+      "model.language_model.layers.4.linear_attn.in_proj_b",
+      "model.language_model.layers.4.linear_attn.in_proj_a",
+      "model.language_model.layers.5.linear_attn",
+      "model.language_model.layers.5.linear_attn.norm",
+      "model.language_model.layers.5.linear_attn.in_proj_b",
+      "model.language_model.layers.5.linear_attn.in_proj_a",
+      "model.language_model.layers.6.linear_attn",
+      "model.language_model.layers.6.linear_attn.norm",
+      "model.language_model.layers.6.linear_attn.in_proj_b",
+      "model.language_model.layers.6.linear_attn.in_proj_a",
+      "model.language_model.layers.8.linear_attn",
+      "model.language_model.layers.8.linear_attn.norm",
+      "model.language_model.layers.8.linear_attn.in_proj_b",
+      "model.language_model.layers.8.linear_attn.in_proj_a",
+      "model.language_model.layers.9.linear_attn",
+      "model.language_model.layers.9.linear_attn.norm",
+      "model.language_model.layers.9.linear_attn.in_proj_b",
+      "model.language_model.layers.9.linear_attn.in_proj_a",
+      "model.language_model.layers.10.linear_attn",
+      "model.language_model.layers.10.linear_attn.norm",
+      "model.language_model.layers.10.linear_attn.in_proj_b",
+      "model.language_model.layers.10.linear_attn.in_proj_a",
+      "model.language_model.layers.12.linear_attn",
+      "model.language_model.layers.12.linear_attn.norm",
+      "model.language_model.layers.12.linear_attn.in_proj_b",
+      "model.language_model.layers.12.linear_attn.in_proj_a",
+      "model.language_model.layers.13.linear_attn",
+      "model.language_model.layers.13.linear_attn.norm",
+      "model.language_model.layers.13.linear_attn.in_proj_b",
+      "model.language_model.layers.13.linear_attn.in_proj_a",
+      "model.language_model.layers.14.linear_attn",
+      "model.language_model.layers.14.linear_attn.norm",
+      "model.language_model.layers.14.linear_attn.in_proj_b",
+      "model.language_model.layers.14.linear_attn.in_proj_a",
+      "model.language_model.layers.16.linear_attn",
+      "model.language_model.layers.16.linear_attn.norm",
+      "model.language_model.layers.16.linear_attn.in_proj_b",
+      "model.language_model.layers.16.linear_attn.in_proj_a",
+      "model.language_model.layers.17.linear_attn",
+      "model.language_model.layers.17.linear_attn.norm",
+      "model.language_model.layers.17.linear_attn.in_proj_b",
+      "model.language_model.layers.17.linear_attn.in_proj_a",
+      "model.language_model.layers.18.linear_attn",
+      "model.language_model.layers.18.linear_attn.norm",
+      "model.language_model.layers.18.linear_attn.in_proj_b",
+      "model.language_model.layers.18.linear_attn.in_proj_a",
+      "model.language_model.layers.20.linear_attn",
+      "model.language_model.layers.20.linear_attn.norm",
+      "model.language_model.layers.20.linear_attn.in_proj_b",
+      "model.language_model.layers.20.linear_attn.in_proj_a",
+      "model.language_model.layers.21.linear_attn",
+      "model.language_model.layers.21.linear_attn.norm",
+      "model.language_model.layers.21.linear_attn.in_proj_b",
+      "model.language_model.layers.21.linear_attn.in_proj_a",
+      "model.language_model.layers.22.linear_attn",
+      "model.language_model.layers.22.linear_attn.norm",
+      "model.language_model.layers.22.linear_attn.in_proj_b",
+      "model.language_model.layers.22.linear_attn.in_proj_a",
+      "model.language_model.layers.24.linear_attn",
+      "model.language_model.layers.24.linear_attn.norm",
+      "model.language_model.layers.24.linear_attn.in_proj_b",
+      "model.language_model.layers.24.linear_attn.in_proj_a",
+      "model.language_model.layers.25.linear_attn",
+      "model.language_model.layers.25.linear_attn.norm",
+      "model.language_model.layers.25.linear_attn.in_proj_b",
+      "model.language_model.layers.25.linear_attn.in_proj_a",
+      "model.language_model.layers.26.linear_attn",
+      "model.language_model.layers.26.linear_attn.norm",
+      "model.language_model.layers.26.linear_attn.in_proj_b",
+      "model.language_model.layers.26.linear_attn.in_proj_a",
+      "model.language_model.layers.28.linear_attn",
+      "model.language_model.layers.28.linear_attn.norm",
+      "model.language_model.layers.28.linear_attn.in_proj_b",
+      "model.language_model.layers.28.linear_attn.in_proj_a",
+      "model.language_model.layers.29.linear_attn",
+      "model.language_model.layers.29.linear_attn.norm",
+      "model.language_model.layers.29.linear_attn.in_proj_b",
+      "model.language_model.layers.29.linear_attn.in_proj_a",
+      "model.language_model.layers.30.linear_attn",
+      "model.language_model.layers.30.linear_attn.norm",
+      "model.language_model.layers.30.linear_attn.in_proj_b",
+      "model.language_model.layers.30.linear_attn.in_proj_a",
+      "model.language_model.layers.32.linear_attn",
+      "model.language_model.layers.32.linear_attn.norm",
+      "model.language_model.layers.32.linear_attn.in_proj_b",
+      "model.language_model.layers.32.linear_attn.in_proj_a",
+      "model.language_model.layers.33.linear_attn",
+      "model.language_model.layers.33.linear_attn.norm",
+      "model.language_model.layers.33.linear_attn.in_proj_b",
+      "model.language_model.layers.33.linear_attn.in_proj_a",
+      "model.language_model.layers.34.linear_attn",
+      "model.language_model.layers.34.linear_attn.norm",
+      "model.language_model.layers.34.linear_attn.in_proj_b",
+      "model.language_model.layers.34.linear_attn.in_proj_a",
+      "model.language_model.layers.36.linear_attn",
+      "model.language_model.layers.36.linear_attn.norm",
+      "model.language_model.layers.36.linear_attn.in_proj_b",
+      "model.language_model.layers.36.linear_attn.in_proj_a",
+      "model.language_model.layers.37.linear_attn",
+      "model.language_model.layers.37.linear_attn.norm",
+      "model.language_model.layers.37.linear_attn.in_proj_b",
+      "model.language_model.layers.37.linear_attn.in_proj_a",
+      "model.language_model.layers.38.linear_attn",
+      "model.language_model.layers.38.linear_attn.norm",
+      "model.language_model.layers.38.linear_attn.in_proj_b",
+      "model.language_model.layers.38.linear_attn.in_proj_a",
+      "model.language_model.layers.40.linear_attn",
+      "model.language_model.layers.40.linear_attn.norm",
+      "model.language_model.layers.40.linear_attn.in_proj_b",
+      "model.language_model.layers.40.linear_attn.in_proj_a",
+      "model.language_model.layers.41.linear_attn",
+      "model.language_model.layers.41.linear_attn.norm",
+      "model.language_model.layers.41.linear_attn.in_proj_b",
+      "model.language_model.layers.41.linear_attn.in_proj_a",
+      "model.language_model.layers.42.linear_attn",
+      "model.language_model.layers.42.linear_attn.norm",
+      "model.language_model.layers.42.linear_attn.in_proj_b",
+      "model.language_model.layers.42.linear_attn.in_proj_a",
+      "model.language_model.layers.44.linear_attn",
+      "model.language_model.layers.44.linear_attn.norm",
+      "model.language_model.layers.44.linear_attn.in_proj_b",
+      "model.language_model.layers.44.linear_attn.in_proj_a",
+      "model.language_model.layers.45.linear_attn",
+      "model.language_model.layers.45.linear_attn.norm",
+      "model.language_model.layers.45.linear_attn.in_proj_b",
+      "model.language_model.layers.45.linear_attn.in_proj_a",
+      "model.language_model.layers.46.linear_attn",
+      "model.language_model.layers.46.linear_attn.norm",
+      "model.language_model.layers.46.linear_attn.in_proj_b",
+      "model.language_model.layers.46.linear_attn.in_proj_a",
+      "model.language_model.layers.48.linear_attn",
+      "model.language_model.layers.48.linear_attn.norm",
+      "model.language_model.layers.48.linear_attn.in_proj_b",
+      "model.language_model.layers.48.linear_attn.in_proj_a",
+      "model.language_model.layers.49.linear_attn",
+      "model.language_model.layers.49.linear_attn.norm",
+      "model.language_model.layers.49.linear_attn.in_proj_b",
+      "model.language_model.layers.49.linear_attn.in_proj_a",
+      "model.language_model.layers.50.linear_attn",
+      "model.language_model.layers.50.linear_attn.norm",
+      "model.language_model.layers.50.linear_attn.in_proj_b",
+      "model.language_model.layers.50.linear_attn.in_proj_a",
+      "model.language_model.layers.52.linear_attn",
+      "model.language_model.layers.52.linear_attn.norm",
+      "model.language_model.layers.52.linear_attn.in_proj_b",
+      "model.language_model.layers.52.linear_attn.in_proj_a",
+      "model.language_model.layers.53.linear_attn",
+      "model.language_model.layers.53.linear_attn.norm",
+      "model.language_model.layers.53.linear_attn.in_proj_b",
+      "model.language_model.layers.53.linear_attn.in_proj_a",
+      "model.language_model.layers.54.linear_attn",
+      "model.language_model.layers.54.linear_attn.norm",
+      "model.language_model.layers.54.linear_attn.in_proj_b",
+      "model.language_model.layers.54.linear_attn.in_proj_a",
+      "model.language_model.layers.56.linear_attn",
+      "model.language_model.layers.56.linear_attn.norm",
+      "model.language_model.layers.56.linear_attn.in_proj_b",
+      "model.language_model.layers.56.linear_attn.in_proj_a",
+      "model.language_model.layers.57.linear_attn",
+      "model.language_model.layers.57.linear_attn.norm",
+      "model.language_model.layers.57.linear_attn.in_proj_b",
+      "model.language_model.layers.57.linear_attn.in_proj_a",
+      "model.language_model.layers.58.linear_attn",
+      "model.language_model.layers.58.linear_attn.norm",
+      "model.language_model.layers.58.linear_attn.in_proj_b",
+      "model.language_model.layers.58.linear_attn.in_proj_a",
+      "model.language_model.layers.60.linear_attn",
+      "model.language_model.layers.60.linear_attn.norm",
+      "model.language_model.layers.60.linear_attn.in_proj_b",
+      "model.language_model.layers.60.linear_attn.in_proj_a",
+      "model.language_model.layers.61.linear_attn",
+      "model.language_model.layers.61.linear_attn.norm",
+      "model.language_model.layers.61.linear_attn.in_proj_b",
+      "model.language_model.layers.61.linear_attn.in_proj_a",
+      "model.language_model.layers.62.linear_attn",
+      "model.language_model.layers.62.linear_attn.norm",
+      "model.language_model.layers.62.linear_attn.in_proj_b",
+      "model.language_model.layers.62.linear_attn.in_proj_a",
+      "lm_head"
+    ],
+    "kv_cache_scheme": null,
+    "quant_method": "compressed-tensors",
+    "quantization_status": "compressed",
+    "sparsity_config": {},
+    "transform_config": {},
+    "version": "0.18.0"
+  },
+  "text_config": {
+    "attention_bias": false,
+    "attention_dropout": 0.0,
+    "attn_output_gate": true,
+    "bos_token_id": 248044,
+    "dtype": "bfloat16",
+    "eos_token_id": 248044,
+    "full_attention_interval": 4,
+    "head_dim": 256,
+    "hidden_act": "silu",
+    "hidden_size": 5120,
+    "initializer_range": 0.02,
+    "intermediate_size": 17408,
+    "layer_types": [
+      "linear_attention",
+      "linear_attention",
+      "linear_attention",
+      "full_attention",
+      "linear_attention",
+      "linear_attention",
+      "linear_attention",
+      "full_attention",
+      "linear_attention",
+      "linear_attention",
+      "linear_attention",
+      "full_attention",
+      "linear_attention",
+      "linear_attention",
+      "linear_attention",
+      "full_attention",
+      "linear_attention",
+      "linear_attention",
+      "linear_attention",
+      "full_attention",
+      "linear_attention",
+      "linear_attention",
+      "linear_attention",
+      "full_attention",
+      "linear_attention",
+      "linear_attention",
+      "linear_attention",
+      "full_attention",
+      "linear_attention",
+      "linear_attention",
+      "linear_attention",
+      "full_attention",
+      "linear_attention",
+      "linear_attention",
+      "linear_attention",
+      "full_attention",
+      "linear_attention",
+      "linear_attention",
+      "linear_attention",
+      "full_attention",
+      "linear_attention",
+      "linear_attention",
+      "linear_attention",
+      "full_attention",
+      "linear_attention",
+      "linear_attention",
+      "linear_attention",
+      "full_attention",
+      "linear_attention",
+      "linear_attention",
+      "linear_attention",
+      "full_attention",
+      "linear_attention",
+      "linear_attention",
+      "linear_attention",
+      "full_attention",
+      "linear_attention",
+      "linear_attention",
+      "linear_attention",
+      "full_attention",
+      "linear_attention",
+      "linear_attention",
+      "linear_attention",
+      "full_attention"
+    ],
+    "linear_conv_kernel_dim": 4,
+    "linear_key_head_dim": 128,
+    "linear_num_key_heads": 16,
+    "linear_num_value_heads": 48,
+    "linear_value_head_dim": 128,
+    "mamba_ssm_dtype": "float32",
+    "max_position_embeddings": 262144,
+    "model_type": "qwen3_5_text",
+    "mtp_num_hidden_layers": 1,
+    "mtp_use_dedicated_embeddings": false,
+    "num_attention_heads": 24,
+    "num_hidden_layers": 64,
+    "num_key_value_heads": 4,
+    "output_gate_type": "swish",
+    "pad_token_id": null,
+    "partial_rotary_factor": 0.25,
+    "rms_norm_eps": 1e-06,
+    "rope_parameters": {
+      "mrope_interleaved": true,
+      "mrope_section": [
+        11,
+        11,
+        10
+      ],
+      "partial_rotary_factor": 0.25,
+      "rope_theta": 10000000,
+      "rope_type": "default"
+    },
+    "tie_word_embeddings": false,
+    "use_cache": true,
+    "vocab_size": 248320
+  },
+  "tie_word_embeddings": false,
+  "transformers_version": "5.14.1",
+  "video_token_id": 248057,
+  "vision_config": {
+    "deepstack_visual_indexes": [],
+    "depth": 27,
+    "dtype": "bfloat16",
+    "hidden_act": "gelu_pytorch_tanh",
+    "hidden_size": 1152,
+    "in_channels": 3,
+    "initializer_range": 0.02,
+    "intermediate_size": 4304,
+    "model_type": "qwen3_5_vision",
+    "num_heads": 16,
+    "num_position_embeddings": 2304,
+    "out_hidden_size": 5120,
+    "patch_size": 16,
+    "spatial_merge_size": 2,
+    "temporal_patch_size": 2
+  },
+  "vision_end_token_id": 248054,
+  "vision_start_token_id": 248053
+}

generation_config.json

@@ -0,0 +1,12 @@
+{
+  "bos_token_id": 248044,
+  "do_sample": true,
+  "eos_token_id": [
+    248046
+  ],
+  "pad_token_id": 248044,
+  "temperature": 1.0,
+  "top_k": 20,
+  "top_p": 0.95,
+  "transformers_version": "5.14.1"
+}

model-mtp.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90fa0e3eed5a647c035c6df9ecabc416c0f8d573ff84ac12485b085f00a7cdf2
+size 849400424

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1adf9ff5b3ce8b3566b2d89f3b70115b043904b05a8ac699a13424b6cfedc90b
+size 18698467264

preprocessor_config.json

@@ -0,0 +1,21 @@
+{
+    "size": {
+        "longest_edge": 16777216,
+        "shortest_edge": 65536
+    },
+    "patch_size": 16,
+    "temporal_patch_size": 2,
+    "merge_size": 2,
+    "image_mean": [
+        0.5,
+        0.5,
+        0.5
+    ],
+    "image_std": [
+        0.5,
+        0.5,
+        0.5
+    ],
+    "processor_class": "Qwen3VLProcessor",
+    "image_processor_type": "Qwen2VLImageProcessorFast"
+}

processor_config.json

@@ -0,0 +1,60 @@
+{
+  "image_processor": {
+    "do_convert_rgb": true,
+    "do_normalize": true,
+    "do_rescale": true,
+    "do_resize": true,
+    "image_mean": [
+      0.5,
+      0.5,
+      0.5
+    ],
+    "image_processor_type": "Qwen2VLImageProcessor",
+    "image_std": [
+      0.5,
+      0.5,
+      0.5
+    ],
+    "merge_size": 2,
+    "patch_size": 16,
+    "resample": 3,
+    "rescale_factor": 0.00392156862745098,
+    "size": {
+      "longest_edge": 16777216,
+      "shortest_edge": 65536
+    },
+    "temporal_patch_size": 2
+  },
+  "processor_class": "Qwen3VLProcessor",
+  "video_processor": {
+    "do_convert_rgb": true,
+    "do_normalize": true,
+    "do_rescale": true,
+    "do_resize": true,
+    "do_sample_frames": true,
+    "fps": 2,
+    "image_mean": [
+      0.5,
+      0.5,
+      0.5
+    ],
+    "image_std": [
+      0.5,
+      0.5,
+      0.5
+    ],
+    "max_frames": 768,
+    "merge_size": 2,
+    "min_frames": 4,
+    "patch_size": 16,
+    "resample": 3,
+    "rescale_factor": 0.00392156862745098,
+    "return_metadata": false,
+    "size": {
+      "longest_edge": 25165824,
+      "shortest_edge": 4096
+    },
+    "temporal_patch_size": 2,
+    "video_processor_type": "Qwen3VLVideoProcessor"
+  }
+}

recipe.yaml

@@ -0,0 +1,605 @@
+default_stage:
+  default_modifiers:
+    AWQModifier:
+      requires_calibration_data: true
+      mappings:
+      - smooth_layer: re:.*layers\.0\.input_layernorm$
+        balance_layers: ['re:.*layers\.0\.linear_attn\.in_proj_qkv$', 're:.*layers\.0\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.0\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.0\.mlp\.gate_proj$', 're:.*layers\.0\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.0\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.0\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.1\.input_layernorm$
+        balance_layers: ['re:.*layers\.1\.linear_attn\.in_proj_qkv$', 're:.*layers\.1\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.1\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.1\.mlp\.gate_proj$', 're:.*layers\.1\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.1\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.1\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.2\.input_layernorm$
+        balance_layers: ['re:.*layers\.2\.linear_attn\.in_proj_qkv$', 're:.*layers\.2\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.2\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.2\.mlp\.gate_proj$', 're:.*layers\.2\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.2\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.2\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.3\.input_layernorm$
+        balance_layers: ['re:.*layers\.3\.self_attn\.q_proj$', 're:.*layers\.3\.self_attn\.k_proj$',
+          're:.*layers\.3\.self_attn\.v_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.3\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.3\.mlp\.gate_proj$', 're:.*layers\.3\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.3\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.3\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.4\.input_layernorm$
+        balance_layers: ['re:.*layers\.4\.linear_attn\.in_proj_qkv$', 're:.*layers\.4\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.4\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.4\.mlp\.gate_proj$', 're:.*layers\.4\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.4\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.4\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.5\.input_layernorm$
+        balance_layers: ['re:.*layers\.5\.linear_attn\.in_proj_qkv$', 're:.*layers\.5\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.5\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.5\.mlp\.gate_proj$', 're:.*layers\.5\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.5\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.5\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.6\.input_layernorm$
+        balance_layers: ['re:.*layers\.6\.linear_attn\.in_proj_qkv$', 're:.*layers\.6\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.6\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.6\.mlp\.gate_proj$', 're:.*layers\.6\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.6\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.6\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.7\.input_layernorm$
+        balance_layers: ['re:.*layers\.7\.self_attn\.q_proj$', 're:.*layers\.7\.self_attn\.k_proj$',
+          're:.*layers\.7\.self_attn\.v_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.7\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.7\.mlp\.gate_proj$', 're:.*layers\.7\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.7\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.7\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.8\.input_layernorm$
+        balance_layers: ['re:.*layers\.8\.linear_attn\.in_proj_qkv$', 're:.*layers\.8\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.8\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.8\.mlp\.gate_proj$', 're:.*layers\.8\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.8\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.8\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.9\.input_layernorm$
+        balance_layers: ['re:.*layers\.9\.linear_attn\.in_proj_qkv$', 're:.*layers\.9\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.9\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.9\.mlp\.gate_proj$', 're:.*layers\.9\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.9\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.9\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.10\.input_layernorm$
+        balance_layers: ['re:.*layers\.10\.linear_attn\.in_proj_qkv$', 're:.*layers\.10\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.10\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.10\.mlp\.gate_proj$', 're:.*layers\.10\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.10\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.10\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.11\.input_layernorm$
+        balance_layers: ['re:.*layers\.11\.self_attn\.q_proj$', 're:.*layers\.11\.self_attn\.k_proj$',
+          're:.*layers\.11\.self_attn\.v_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.11\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.11\.mlp\.gate_proj$', 're:.*layers\.11\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.11\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.11\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.12\.input_layernorm$
+        balance_layers: ['re:.*layers\.12\.linear_attn\.in_proj_qkv$', 're:.*layers\.12\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.12\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.12\.mlp\.gate_proj$', 're:.*layers\.12\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.12\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.12\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.13\.input_layernorm$
+        balance_layers: ['re:.*layers\.13\.linear_attn\.in_proj_qkv$', 're:.*layers\.13\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.13\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.13\.mlp\.gate_proj$', 're:.*layers\.13\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.13\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.13\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.14\.input_layernorm$
+        balance_layers: ['re:.*layers\.14\.linear_attn\.in_proj_qkv$', 're:.*layers\.14\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.14\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.14\.mlp\.gate_proj$', 're:.*layers\.14\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.14\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.14\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.15\.input_layernorm$
+        balance_layers: ['re:.*layers\.15\.self_attn\.q_proj$', 're:.*layers\.15\.self_attn\.k_proj$',
+          're:.*layers\.15\.self_attn\.v_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.15\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.15\.mlp\.gate_proj$', 're:.*layers\.15\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.15\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.15\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.16\.input_layernorm$
+        balance_layers: ['re:.*layers\.16\.linear_attn\.in_proj_qkv$', 're:.*layers\.16\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.16\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.16\.mlp\.gate_proj$', 're:.*layers\.16\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.16\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.16\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.17\.input_layernorm$
+        balance_layers: ['re:.*layers\.17\.linear_attn\.in_proj_qkv$', 're:.*layers\.17\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.17\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.17\.mlp\.gate_proj$', 're:.*layers\.17\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.17\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.17\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.18\.input_layernorm$
+        balance_layers: ['re:.*layers\.18\.linear_attn\.in_proj_qkv$', 're:.*layers\.18\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.18\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.18\.mlp\.gate_proj$', 're:.*layers\.18\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.18\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.18\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.19\.input_layernorm$
+        balance_layers: ['re:.*layers\.19\.self_attn\.q_proj$', 're:.*layers\.19\.self_attn\.k_proj$',
+          're:.*layers\.19\.self_attn\.v_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.19\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.19\.mlp\.gate_proj$', 're:.*layers\.19\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.19\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.19\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.20\.input_layernorm$
+        balance_layers: ['re:.*layers\.20\.linear_attn\.in_proj_qkv$', 're:.*layers\.20\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.20\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.20\.mlp\.gate_proj$', 're:.*layers\.20\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.20\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.20\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.21\.input_layernorm$
+        balance_layers: ['re:.*layers\.21\.linear_attn\.in_proj_qkv$', 're:.*layers\.21\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.21\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.21\.mlp\.gate_proj$', 're:.*layers\.21\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.21\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.21\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.22\.input_layernorm$
+        balance_layers: ['re:.*layers\.22\.linear_attn\.in_proj_qkv$', 're:.*layers\.22\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.22\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.22\.mlp\.gate_proj$', 're:.*layers\.22\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.22\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.22\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.23\.input_layernorm$
+        balance_layers: ['re:.*layers\.23\.self_attn\.q_proj$', 're:.*layers\.23\.self_attn\.k_proj$',
+          're:.*layers\.23\.self_attn\.v_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.23\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.23\.mlp\.gate_proj$', 're:.*layers\.23\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.23\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.23\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.24\.input_layernorm$
+        balance_layers: ['re:.*layers\.24\.linear_attn\.in_proj_qkv$', 're:.*layers\.24\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.24\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.24\.mlp\.gate_proj$', 're:.*layers\.24\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.24\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.24\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.25\.input_layernorm$
+        balance_layers: ['re:.*layers\.25\.linear_attn\.in_proj_qkv$', 're:.*layers\.25\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.25\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.25\.mlp\.gate_proj$', 're:.*layers\.25\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.25\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.25\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.26\.input_layernorm$
+        balance_layers: ['re:.*layers\.26\.linear_attn\.in_proj_qkv$', 're:.*layers\.26\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.26\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.26\.mlp\.gate_proj$', 're:.*layers\.26\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.26\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.26\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.27\.input_layernorm$
+        balance_layers: ['re:.*layers\.27\.self_attn\.q_proj$', 're:.*layers\.27\.self_attn\.k_proj$',
+          're:.*layers\.27\.self_attn\.v_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.27\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.27\.mlp\.gate_proj$', 're:.*layers\.27\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.27\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.27\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.28\.input_layernorm$
+        balance_layers: ['re:.*layers\.28\.linear_attn\.in_proj_qkv$', 're:.*layers\.28\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.28\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.28\.mlp\.gate_proj$', 're:.*layers\.28\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.28\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.28\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.29\.input_layernorm$
+        balance_layers: ['re:.*layers\.29\.linear_attn\.in_proj_qkv$', 're:.*layers\.29\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.29\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.29\.mlp\.gate_proj$', 're:.*layers\.29\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.29\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.29\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.30\.input_layernorm$
+        balance_layers: ['re:.*layers\.30\.linear_attn\.in_proj_qkv$', 're:.*layers\.30\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.30\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.30\.mlp\.gate_proj$', 're:.*layers\.30\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.30\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.30\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.31\.input_layernorm$
+        balance_layers: ['re:.*layers\.31\.self_attn\.q_proj$', 're:.*layers\.31\.self_attn\.k_proj$',
+          're:.*layers\.31\.self_attn\.v_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.31\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.31\.mlp\.gate_proj$', 're:.*layers\.31\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.31\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.31\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.32\.input_layernorm$
+        balance_layers: ['re:.*layers\.32\.linear_attn\.in_proj_qkv$', 're:.*layers\.32\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.32\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.32\.mlp\.gate_proj$', 're:.*layers\.32\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.32\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.32\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.33\.input_layernorm$
+        balance_layers: ['re:.*layers\.33\.linear_attn\.in_proj_qkv$', 're:.*layers\.33\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.33\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.33\.mlp\.gate_proj$', 're:.*layers\.33\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.33\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.33\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.34\.input_layernorm$
+        balance_layers: ['re:.*layers\.34\.linear_attn\.in_proj_qkv$', 're:.*layers\.34\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.34\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.34\.mlp\.gate_proj$', 're:.*layers\.34\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.34\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.34\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.35\.input_layernorm$
+        balance_layers: ['re:.*layers\.35\.self_attn\.q_proj$', 're:.*layers\.35\.self_attn\.k_proj$',
+          're:.*layers\.35\.self_attn\.v_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.35\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.35\.mlp\.gate_proj$', 're:.*layers\.35\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.35\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.35\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.36\.input_layernorm$
+        balance_layers: ['re:.*layers\.36\.linear_attn\.in_proj_qkv$', 're:.*layers\.36\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.36\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.36\.mlp\.gate_proj$', 're:.*layers\.36\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.36\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.36\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.37\.input_layernorm$
+        balance_layers: ['re:.*layers\.37\.linear_attn\.in_proj_qkv$', 're:.*layers\.37\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.37\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.37\.mlp\.gate_proj$', 're:.*layers\.37\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.37\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.37\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.38\.input_layernorm$
+        balance_layers: ['re:.*layers\.38\.linear_attn\.in_proj_qkv$', 're:.*layers\.38\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.38\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.38\.mlp\.gate_proj$', 're:.*layers\.38\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.38\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.38\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.39\.input_layernorm$
+        balance_layers: ['re:.*layers\.39\.self_attn\.q_proj$', 're:.*layers\.39\.self_attn\.k_proj$',
+          're:.*layers\.39\.self_attn\.v_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.39\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.39\.mlp\.gate_proj$', 're:.*layers\.39\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.39\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.39\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.40\.input_layernorm$
+        balance_layers: ['re:.*layers\.40\.linear_attn\.in_proj_qkv$', 're:.*layers\.40\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.40\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.40\.mlp\.gate_proj$', 're:.*layers\.40\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.40\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.40\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.41\.input_layernorm$
+        balance_layers: ['re:.*layers\.41\.linear_attn\.in_proj_qkv$', 're:.*layers\.41\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.41\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.41\.mlp\.gate_proj$', 're:.*layers\.41\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.41\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.41\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.42\.input_layernorm$
+        balance_layers: ['re:.*layers\.42\.linear_attn\.in_proj_qkv$', 're:.*layers\.42\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.42\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.42\.mlp\.gate_proj$', 're:.*layers\.42\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.42\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.42\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.43\.input_layernorm$
+        balance_layers: ['re:.*layers\.43\.self_attn\.q_proj$', 're:.*layers\.43\.self_attn\.k_proj$',
+          're:.*layers\.43\.self_attn\.v_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.43\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.43\.mlp\.gate_proj$', 're:.*layers\.43\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.43\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.43\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.44\.input_layernorm$
+        balance_layers: ['re:.*layers\.44\.linear_attn\.in_proj_qkv$', 're:.*layers\.44\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.44\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.44\.mlp\.gate_proj$', 're:.*layers\.44\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.44\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.44\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.45\.input_layernorm$
+        balance_layers: ['re:.*layers\.45\.linear_attn\.in_proj_qkv$', 're:.*layers\.45\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.45\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.45\.mlp\.gate_proj$', 're:.*layers\.45\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.45\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.45\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.46\.input_layernorm$
+        balance_layers: ['re:.*layers\.46\.linear_attn\.in_proj_qkv$', 're:.*layers\.46\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.46\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.46\.mlp\.gate_proj$', 're:.*layers\.46\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.46\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.46\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.47\.input_layernorm$
+        balance_layers: ['re:.*layers\.47\.self_attn\.q_proj$', 're:.*layers\.47\.self_attn\.k_proj$',
+          're:.*layers\.47\.self_attn\.v_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.47\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.47\.mlp\.gate_proj$', 're:.*layers\.47\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.47\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.47\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.48\.input_layernorm$
+        balance_layers: ['re:.*layers\.48\.linear_attn\.in_proj_qkv$', 're:.*layers\.48\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.48\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.48\.mlp\.gate_proj$', 're:.*layers\.48\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.48\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.48\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.49\.input_layernorm$
+        balance_layers: ['re:.*layers\.49\.linear_attn\.in_proj_qkv$', 're:.*layers\.49\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.49\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.49\.mlp\.gate_proj$', 're:.*layers\.49\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.49\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.49\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.50\.input_layernorm$
+        balance_layers: ['re:.*layers\.50\.linear_attn\.in_proj_qkv$', 're:.*layers\.50\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.50\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.50\.mlp\.gate_proj$', 're:.*layers\.50\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.50\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.50\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.51\.input_layernorm$
+        balance_layers: ['re:.*layers\.51\.self_attn\.q_proj$', 're:.*layers\.51\.self_attn\.k_proj$',
+          're:.*layers\.51\.self_attn\.v_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.51\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.51\.mlp\.gate_proj$', 're:.*layers\.51\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.51\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.51\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.52\.input_layernorm$
+        balance_layers: ['re:.*layers\.52\.linear_attn\.in_proj_qkv$', 're:.*layers\.52\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.52\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.52\.mlp\.gate_proj$', 're:.*layers\.52\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.52\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.52\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.53\.input_layernorm$
+        balance_layers: ['re:.*layers\.53\.linear_attn\.in_proj_qkv$', 're:.*layers\.53\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.53\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.53\.mlp\.gate_proj$', 're:.*layers\.53\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.53\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.53\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.54\.input_layernorm$
+        balance_layers: ['re:.*layers\.54\.linear_attn\.in_proj_qkv$', 're:.*layers\.54\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.54\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.54\.mlp\.gate_proj$', 're:.*layers\.54\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.54\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.54\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.55\.input_layernorm$
+        balance_layers: ['re:.*layers\.55\.self_attn\.q_proj$', 're:.*layers\.55\.self_attn\.k_proj$',
+          're:.*layers\.55\.self_attn\.v_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.55\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.55\.mlp\.gate_proj$', 're:.*layers\.55\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.55\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.55\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.56\.input_layernorm$
+        balance_layers: ['re:.*layers\.56\.linear_attn\.in_proj_qkv$', 're:.*layers\.56\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.56\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.56\.mlp\.gate_proj$', 're:.*layers\.56\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.56\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.56\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.57\.input_layernorm$
+        balance_layers: ['re:.*layers\.57\.linear_attn\.in_proj_qkv$', 're:.*layers\.57\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.57\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.57\.mlp\.gate_proj$', 're:.*layers\.57\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.57\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.57\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.58\.input_layernorm$
+        balance_layers: ['re:.*layers\.58\.linear_attn\.in_proj_qkv$', 're:.*layers\.58\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.58\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.58\.mlp\.gate_proj$', 're:.*layers\.58\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.58\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.58\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.59\.input_layernorm$
+        balance_layers: ['re:.*layers\.59\.self_attn\.q_proj$', 're:.*layers\.59\.self_attn\.k_proj$',
+          're:.*layers\.59\.self_attn\.v_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.59\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.59\.mlp\.gate_proj$', 're:.*layers\.59\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.59\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.59\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.60\.input_layernorm$
+        balance_layers: ['re:.*layers\.60\.linear_attn\.in_proj_qkv$', 're:.*layers\.60\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.60\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.60\.mlp\.gate_proj$', 're:.*layers\.60\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.60\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.60\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.61\.input_layernorm$
+        balance_layers: ['re:.*layers\.61\.linear_attn\.in_proj_qkv$', 're:.*layers\.61\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.61\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.61\.mlp\.gate_proj$', 're:.*layers\.61\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.61\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.61\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.62\.input_layernorm$
+        balance_layers: ['re:.*layers\.62\.linear_attn\.in_proj_qkv$', 're:.*layers\.62\.linear_attn\.in_proj_z$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.62\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.62\.mlp\.gate_proj$', 're:.*layers\.62\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.62\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.62\.mlp\.down_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.63\.input_layernorm$
+        balance_layers: ['re:.*layers\.63\.self_attn\.q_proj$', 're:.*layers\.63\.self_attn\.k_proj$',
+          're:.*layers\.63\.self_attn\.v_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.63\.post_attention_layernorm$
+        balance_layers: ['re:.*layers\.63\.mlp\.gate_proj$', 're:.*layers\.63\.mlp\.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.63\.mlp\.up_proj$
+        balance_layers: ['re:.*layers\.63\.mlp\.down_proj$']
+        activation_hook_target: null
+      duo_scaling: true
+      n_grid: 20
+    QuantizationModifier:
+      targets: [Linear]
+      ignore: [lm_head, 're:.*visual.*', 're:mtp\..*', 're:.*in_proj_a$', 're:.*in_proj_b$']
+      scheme: W4A16_ASYM
+      bypass_divisibility_checks: false
+      requires_calibration_data: false

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:06b9509352d2af50381ab2247e083b80d32d5c0aba91c272ca9ff729b6a0e523
+size 19989325

tokenizer_config.json

@@ -0,0 +1,32 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": true,
+  "local_files_only": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}

video_preprocessor_config.json

@@ -0,0 +1,21 @@
+{
+    "size": {
+        "longest_edge": 25165824,
+        "shortest_edge": 4096
+    },
+    "patch_size": 16,
+    "temporal_patch_size": 2,
+    "merge_size": 2,
+    "image_mean": [
+        0.5,
+        0.5,
+        0.5
+    ],
+    "image_std": [
+        0.5,
+        0.5,
+        0.5
+    ],
+    "processor_class": "Qwen3VLProcessor",
+    "video_processor_type": "Qwen3VLVideoProcessor"
+}