Upload REAP expert-pruned AWQ model (410/512 experts, W4A16)

.gitattributes

@@ -33,3 +33,5 @@ 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
+model.safetensors.index.json filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,246 @@
+---
+language:
+  - en
+license: other
+license_name: qwen
+license_link: https://huggingface.co/Qwen/Qwen3-Coder-Next/blob/main/LICENSE
+tags:
+  - moe
+  - pruning
+  - awq
+  - quantized
+  - qwen3
+  - reap
+  - expert-pruning
+  - compressed-tensors
+  - 4bit
+  - research
+base_model: Qwen/Qwen3-Coder-Next
+pipeline_tag: text-generation
+library_name: transformers
+model-index:
+  - name: Qwen3-Coder-Next-REAP-AWQ
+    results: []
+---
+
+# Qwen3-Coder-Next-REAP-AWQ
+
+**Expert-pruned and AWQ-quantized Qwen3-Coder-Next.** 20% of MoE experts removed via REAP saliency analysis across diverse calibration data, then quantized to W4A16. The result is a model that runs ~7-12% faster at the token level, uses significantly less VRAM, and frees up memory for larger KV caches and higher concurrency -- at the cost of occasional quality regressions on certain tasks.
+
+> **Status:** Research / Experimental. This model produces usable output across code, math, reasoning, and general tasks, but **some outputs may be less polished than the unpruned baseline** -- particularly around structured output formatting and multi-step logic chains. It works. It's faster. It's not perfect. See [Limitations](#limitations) for specifics.
+
+## Why This Exists
+
+Qwen3-Coder-Next is a large Mixture-of-Experts model with **512 experts per layer**, but only **10 are active per token**. That means ~98% of expert parameters sit idle for any given input. This creates an opportunity: measure which experts matter least across a diverse workload and remove them.
+
+Fewer experts means:
+- **Smaller model** (~32 GB vs ~37 GB for unpruned AWQ) -- 5 GB freed for KV cache
+- **Faster inference** -- less memory bandwidth pressure, fewer expert weights to page
+- **Higher concurrency** -- more VRAM headroom for batched requests
+
+The trade-off is a small quality hit on some tasks, which we document below.
+
+## Model Details
+
+| Property | Value |
+|----------|-------|
+| Base Model | [Qwen/Qwen3-Coder-Next](https://huggingface.co/Qwen/Qwen3-Coder-Next) (BF16, 149 GB) |
+| Architecture | `Qwen3NextForCausalLM` (MoE + Gated DeltaNet hybrid attention) |
+| Layers | 48 (36 linear attention + 12 full attention) |
+| Original Experts | 512 per layer, 10 active per token |
+| After Pruning | **410 per layer** (20% removed via REAP) |
+| Quantization | AWQ W4A16, symmetric, group_size=128 |
+| Format | `compressed-tensors` (compatible with vLLM, transformers) |
+| Context Length | 262,144 tokens |
+| Size on Disk | ~32 GB (7 shards) |
+
+## Evaluation Results
+
+Evaluated on a custom benchmark suite covering code generation, reasoning, tool use, math, general knowledge, and writing. Each test is pass/fail with latency metrics.
+
+### REAP-AWQ (This Model, 410 experts)
+
+| Run | Categories | Pass Rate | Avg tok/s | Notes |
+|-----|-----------|-----------|-----------|-------|
+| Run 1 | code, reasoning, scaffold, chat | **13/17 (76%)** | **141.4** | logic_puzzle, instruction_following, JSON formatting failed |
+| Run 2 | code, math, general, writing | **17/17 (100%)** | **136.0** | Clean sweep on expanded test set |
+
+### Baseline (Unpruned AWQ, 512 experts)
+
+| Run | Categories | Pass Rate | Avg tok/s | Notes |
+|-----|-----------|-----------|-----------|-------|
+| Run 1 | code, reasoning, scaffold, chat | 13/17 (76%) | 126.3 | Tool calling failed (vLLM config issue) |
+| Run 2 | code, reasoning, scaffold, chat | 16/17 (94%) | 132.0 | Stable after vLLM restart |
+
+### Speed Comparison
+
+Across comparable test categories, the REAP model is consistently faster:
+
+| Category | Baseline (tok/s) | REAP (tok/s) | Speedup |
+|----------|-----------------|--------------|---------|
+| Code | 126.2 | 135.1 | +7.1% |
+| Reasoning | 131.5 | 140.9 | +7.2% |
+| Scaffold / Tool Use | 142.3 | 152.1 | +6.9% |
+| Chat | 128.7 | 138.2 | +7.4% |
+
+These numbers reflect single-request latency. In practice, the VRAM savings compound at higher concurrency -- the freed memory supports larger batch sizes and longer contexts, where we've observed up to **~14% effective throughput gains** in multi-request serving scenarios.
+
+## How It Was Made
+
+### The REAP Pipeline
+
+**REAP** (Robust Expert Architecture Pruning) scores each expert by combining activation frequency with contribution magnitude:
+
+```
+REAP(expert) = sum(activation_norm * router_weight) / total_tokens
+```
+
+Experts that rarely fire *and* contribute little when they do get the lowest scores and are pruned first.
+
+### Diverse Calibration
+
+A code model uses different experts for different tasks. Pruning based on code-only data risks removing experts critical for reasoning, creative writing, or math. We calibrated across 4 datasets:
+
+| Dataset | Domain | Purpose |
+|---------|--------|---------|
+| `evol-codealpaca-v1` | Code | Core competency |
+| `allenai/c4` | Web text | General language |
+| `WritingPrompts_curated` | Creative writing | Long-form generation |
+| `tulu-3-sft-personas-math` | Math | Reasoning chains |
+
+256 samples per dataset, merged by summing accumulator metrics before computing derived scores.
+
+### Super-Expert Preservation
+
+Some experts have extremely high peak activations -- they fire rarely but are critical when they do (e.g., handling rare syntax patterns or domain-specific tokens). These "super-experts" are protected from pruning regardless of their average REAP score, preventing catastrophic failures on rare inputs.
+
+### Quantization
+
+After pruning, the remaining 410 experts are quantized using AWQ:
+
+- **Scheme:** W4A16, symmetric, group_size=128
+- **Calibration:** 256 samples from `evol-codealpaca-v1`
+- **Format:** `compressed-tensors` via `llmcompressor`
+
+**Layers kept at full precision** (these are sensitive to quantization):
+- MoE router gates (`mlp.gate`, `mlp.shared_expert_gate`)
+- Gated DeltaNet internals (`linear_attn.conv1d`, `in_proj_a`, `in_proj_b`)
+- Output head (`lm_head`)
+
+### Memory-Managed Execution
+
+The full pipeline ran on **4x RTX 3090 (96 GB VRAM)** with **128 GB system RAM**. The 149 GB BF16 model doesn't fit in GPU memory, so each phase runs as a separate OS process with CPU offload:
+
+1. **Observe** (4 runs, one per dataset) -- hooks accumulate statistics to CPU RAM
+2. **Merge** -- CPU-only, combines observations
+3. **Prune** -- fresh model load, in-place expert removal
+4. **AWQ** -- `max_memory` caps at 20 GiB/GPU with 100 GiB CPU overflow
+
+Process isolation guarantees clean GPU state between phases.
+
+## Limitations
+
+This is an experimental research model. Known issues:
+
+- **Self-correction loops:** On some prompts, the model second-guesses itself more than the baseline ("Wait -- let me re-check..."), producing verbose but ultimately correct answers. This appears to be an artifact of the pruning affecting confidence calibration.
+- **Structured output:** Occasional JSON formatting errors (e.g., missing closing brackets). The model understands the structure but sometimes truncates. Constrained decoding (e.g., vLLM's `guided_json`) mitigates this.
+- **Logic puzzles:** The model struggles with certain ordering/constraint satisfaction problems that the baseline also finds difficult. Pruning didn't help here.
+- **Instruction following edge cases:** Rarely drops minor formatting instructions (e.g., numbered lists vs. unnumbered). Core instruction comprehension is intact.
+
+None of these are showstoppers for most use cases. The model handles code generation, mathematical reasoning, general Q&A, creative writing, and tool calling well.
+
+## Usage
+
+### Serving with vLLM
+
+```bash
+vllm serve mtecnic/Qwen3-Coder-Next-REAP-AWQ \
+    --tensor-parallel-size 4 \
+    --gpu-memory-utilization 0.93 \
+    --trust-remote-code \
+    --max-model-len 32768 \
+    --max-num-seqs 16
+```
+
+### Python (Transformers)
+
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+model = AutoModelForCausalLM.from_pretrained(
+    "mtecnic/Qwen3-Coder-Next-REAP-AWQ",
+    device_map="auto",
+    torch_dtype="auto",
+    trust_remote_code=True,
+)
+tokenizer = AutoTokenizer.from_pretrained("mtecnic/Qwen3-Coder-Next-REAP-AWQ")
+
+messages = [{"role": "user", "content": "Write a Python function to merge two sorted lists."}]
+inputs = tokenizer.apply_chat_template(messages, return_tensors="pt", add_generation_prompt=True)
+outputs = model.generate(inputs.to(model.device), max_new_tokens=512)
+print(tokenizer.decode(outputs[0], skip_special_tokens=True))
+```
+
+## Pipeline Configuration
+
+```yaml
+# Observation
+samples_per_dataset: 256
+max_sequence_length: 512
+distance_metric: cosine
+datasets:
+  - evol-codealpaca-v1
+  - c4
+  - WritingPrompts_curated
+  - tulu-3-sft-personas-math
+
+# Pruning
+method: reap
+compression_ratio: 0.20
+preserve_super_experts: true
+seed: 42
+
+# Quantization
+method: awq
+scheme: W4A16
+group_size: 128
+calibration_samples: 256
+```
+
+## Research Context
+
+This model is the result of extensive experimentation with MoE expert pruning. Key learnings:
+
+- **40% compression was too aggressive** -- an earlier attempt removing 205/512 experts per layer caused noticeable quality degradation across all categories.
+- **20% is the sweet spot for this architecture** -- quality is largely preserved while delivering meaningful speed and memory improvements.
+- **Diverse calibration is essential** -- code-only calibration misidentifies experts that are critical for reasoning and general language tasks.
+- **Super-expert preservation prevents catastrophic edge cases** -- without it, rare but important patterns (unusual syntax, domain-specific tokens) break completely.
+- **The Gated DeltaNet layers are fragile** -- quantizing the linear attention internals (`conv1d`, `in_proj_a/b`) caused significant quality loss. Keeping them at full precision is non-negotiable.
+
+## Hardware Requirements
+
+- **Minimum:** 4x 24 GB GPUs (e.g., RTX 3090/4090) with tensor parallelism
+- **Recommended:** 2x 48 GB GPUs (e.g., A6000) or 1x 80 GB GPU (e.g., A100/H100)
+- The ~5 GB VRAM savings vs. unpruned AWQ is most impactful on memory-constrained setups
+
+## Acknowledgments
+
+- **[Qwen](https://github.com/QwenLM/Qwen3)** for Qwen3-Coder-Next
+- **[Cerebras](https://www.cerebras.net/)** for the REAP pruning methodology
+- **[MIT HAN Lab](https://github.com/mit-han-lab/llm-awq)** for AWQ
+- **[Neural Magic / vLLM](https://github.com/vllm-project/vllm)** for `llmcompressor` and efficient MoE serving
+
+## License
+
+This model inherits the license of the base [Qwen3-Coder-Next](https://huggingface.co/Qwen/Qwen3-Coder-Next) model. See the Qwen license for details.
+
+## Citation
+
+```bibtex
+@misc{wienandt2026reap_awq,
+  title={REAP Expert Pruning of Qwen3-Coder-Next: 20\% Expert Reduction with AWQ Quantization},
+  author={Nic Wienandt},
+  year={2026},
+  url={https://huggingface.co/mtecnic/Qwen3-Coder-Next-REAP-AWQ}
+}
+```

added_tokens.json

@@ -0,0 +1,28 @@
+{
+  "</think>": 151668,
+  "</tool_call>": 151658,
+  "</tool_response>": 151666,
+  "<think>": 151667,
+  "<tool_call>": 151657,
+  "<tool_response>": 151665,
+  "<|box_end|>": 151649,
+  "<|box_start|>": 151648,
+  "<|endoftext|>": 151643,
+  "<|file_sep|>": 151664,
+  "<|fim_middle|>": 151660,
+  "<|fim_pad|>": 151662,
+  "<|fim_prefix|>": 151659,
+  "<|fim_suffix|>": 151661,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644,
+  "<|image_pad|>": 151655,
+  "<|object_ref_end|>": 151647,
+  "<|object_ref_start|>": 151646,
+  "<|quad_end|>": 151651,
+  "<|quad_start|>": 151650,
+  "<|repo_name|>": 151663,
+  "<|video_pad|>": 151656,
+  "<|vision_end|>": 151653,
+  "<|vision_pad|>": 151654,
+  "<|vision_start|>": 151652
+}

chat_template.jinja

@@ -0,0 +1,117 @@
+{% macro render_extra_keys(json_dict, handled_keys) %}
+    {%- if json_dict is mapping %}
+        {%- for json_key in json_dict if json_key not in handled_keys %}
+            {%- if json_dict[json_key] is string %}
+                {{-'\n<' ~ json_key ~ '>' ~ (json_dict[json_key] | string) ~ '</' ~ json_key ~ '>' }}
+            {%- else %}
+                {{- '\n<' ~ json_key ~ '>' ~ (json_dict[json_key] | tojson | safe) ~ '</' ~ json_key ~ '>' }}
+            {%- endif %}
+        {%- endfor %}
+    {%- endif %}
+{%- endmacro %}
+
+{%- if messages[0]["role"] == "system" %}
+    {%- set system_message = messages[0]["content"] %}
+    {%- set loop_messages = messages[1:] %}
+{%- else %}
+    {%- set loop_messages = messages %}
+{%- endif %}
+
+{%- if not tools is defined %}
+    {%- set tools = [] %}
+{%- endif %}
+
+{%- if system_message is defined %}
+    {{- "<|im_start|>system\n" + system_message }}
+{%- else %}
+    {%- if tools is iterable and tools | length > 0 %}
+        {{- "<|im_start|>system\nYou are Qwen, a helpful AI assistant that can interact with a computer to solve tasks." }}
+    {%- endif %}
+{%- endif %}
+{%- if tools is iterable and tools | length > 0 %}
+    {{- "\n\n# Tools\n\nYou have access to the following functions:\n\n" }}
+    {{- "<tools>" }}
+    {%- for tool in tools %}
+        {%- if tool.function is defined %}
+            {%- set tool = tool.function %}
+        {%- endif %}
+        {{- "\n<function>\n<name>" ~ tool.name ~ "</name>" }}
+        {%- if tool.description is defined %}
+            {{- '\n<description>' ~ (tool.description | trim) ~ '</description>' }}
+        {%- endif %}
+        {{- '\n<parameters>' }}
+        {%- if tool.parameters is defined and tool.parameters is mapping and tool.parameters.properties is defined and tool.parameters.properties is mapping %}
+            {%- for param_name, param_fields in tool.parameters.properties|items %}
+                {{- '\n<parameter>' }}
+                {{- '\n<name>' ~ param_name ~ '</name>' }}
+                {%- if param_fields.type is defined %}
+                    {{- '\n<type>' ~ (param_fields.type | string) ~ '</type>' }}
+                {%- endif %}
+                {%- if param_fields.description is defined %}
+                    {{- '\n<description>' ~ (param_fields.description | trim) ~ '</description>' }}
+                {%- endif %}
+                {%- set handled_keys = ['name', 'type', 'description'] %}
+                {{- render_extra_keys(param_fields, handled_keys) }}
+                {{- '\n</parameter>' }}
+            {%- endfor %}
+        {%- endif %}
+        {%- set handled_keys = ['type', 'properties'] %}
+        {{- render_extra_keys(tool.parameters, handled_keys) }}
+        {{- '\n</parameters>' }}
+        {%- set handled_keys = ['type', 'name', 'description', 'parameters'] %}
+        {{- render_extra_keys(tool, handled_keys) }}
+        {{- '\n</function>' }}
+    {%- 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>' }}
+{%- endif %}
+{%- if system_message is defined %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if tools is iterable and tools | length > 0 %}
+        {{- '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- for message in loop_messages %}
+    {%- if message.role == "assistant" and message.tool_calls is defined and message.tool_calls is iterable and message.tool_calls | length > 0 %}
+        {{- '<|im_start|>' + message.role }}
+        {%- if message.content is defined and message.content is string and message.content | trim | length > 0 %}
+            {{- '\n' + message.content | trim + '\n' }}
+        {%- endif %}
+        {%- for tool_call in message.tool_calls %}
+            {%- if tool_call.function is defined %}
+                {%- set tool_call = tool_call.function %}
+            {%- endif %}
+            {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+            {%- if tool_call.arguments is defined %}
+                {%- for args_name, args_value in tool_call.arguments|items %}
+                    {{- '<parameter=' + args_name + '>\n' }}
+                    {%- set args_value = args_value if args_value is string else args_value | tojson | safe %}
+                    {{- args_value }}
+                    {{- '\n</parameter>\n' }}
+                {%- endfor %}
+            {%- endif %}
+            {{- '</function>\n</tool_call>' }}
+        {%- endfor %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "user" or message.role == "system" or message.role == "assistant" %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+{%- endif %}

config.json

@@ -0,0 +1,263 @@
+{
+  "architectures": [
+    "Qwen3NextForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0,
+  "bos_token_id": 151643,
+  "decoder_sparse_step": 1,
+  "dtype": "bfloat16",
+  "eos_token_id": 151645,
+  "full_attention_interval": 4,
+  "head_dim": 256,
+  "hidden_act": "silu",
+  "hidden_size": 2048,
+  "initializer_range": 0.02,
+  "intermediate_size": 5120,
+  "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_conv_kernel_dim": 4,
+  "linear_key_head_dim": 128,
+  "linear_num_key_heads": 16,
+  "linear_num_value_heads": 32,
+  "linear_value_head_dim": 128,
+  "max_position_embeddings": 262144,
+  "mlp_only_layers": [],
+  "model_type": "qwen3_next",
+  "moe_intermediate_size": 512,
+  "norm_topk_prob": true,
+  "num_attention_heads": 16,
+  "num_experts": 410,
+  "num_experts_per_tok": 10,
+  "num_hidden_layers": 48,
+  "num_key_value_heads": 2,
+  "output_router_logits": false,
+  "partial_rotary_factor": 0.25,
+  "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": true,
+          "type": "int",
+          "zp_dtype": null
+        }
+      }
+    },
+    "format": "pack-quantized",
+    "global_compression_ratio": null,
+    "ignore": [
+      "model.layers.0.linear_attn.in_proj_ba",
+      "model.layers.0.mlp.gate",
+      "model.layers.0.mlp.shared_expert_gate",
+      "model.layers.1.linear_attn.in_proj_ba",
+      "model.layers.1.mlp.gate",
+      "model.layers.1.mlp.shared_expert_gate",
+      "model.layers.2.linear_attn.in_proj_ba",
+      "model.layers.2.mlp.gate",
+      "model.layers.2.mlp.shared_expert_gate",
+      "model.layers.3.mlp.gate",
+      "model.layers.3.mlp.shared_expert_gate",
+      "model.layers.4.linear_attn.in_proj_ba",
+      "model.layers.4.mlp.gate",
+      "model.layers.4.mlp.shared_expert_gate",
+      "model.layers.5.linear_attn.in_proj_ba",
+      "model.layers.5.mlp.gate",
+      "model.layers.5.mlp.shared_expert_gate",
+      "model.layers.6.linear_attn.in_proj_ba",
+      "model.layers.6.mlp.gate",
+      "model.layers.6.mlp.shared_expert_gate",
+      "model.layers.7.mlp.gate",
+      "model.layers.7.mlp.shared_expert_gate",
+      "model.layers.8.linear_attn.in_proj_ba",
+      "model.layers.8.mlp.gate",
+      "model.layers.8.mlp.shared_expert_gate",
+      "model.layers.9.linear_attn.in_proj_ba",
+      "model.layers.9.mlp.gate",
+      "model.layers.9.mlp.shared_expert_gate",
+      "model.layers.10.linear_attn.in_proj_ba",
+      "model.layers.10.mlp.gate",
+      "model.layers.10.mlp.shared_expert_gate",
+      "model.layers.11.mlp.gate",
+      "model.layers.11.mlp.shared_expert_gate",
+      "model.layers.12.linear_attn.in_proj_ba",
+      "model.layers.12.mlp.gate",
+      "model.layers.12.mlp.shared_expert_gate",
+      "model.layers.13.linear_attn.in_proj_ba",
+      "model.layers.13.mlp.gate",
+      "model.layers.13.mlp.shared_expert_gate",
+      "model.layers.14.linear_attn.in_proj_ba",
+      "model.layers.14.mlp.gate",
+      "model.layers.14.mlp.shared_expert_gate",
+      "model.layers.15.mlp.gate",
+      "model.layers.15.mlp.shared_expert_gate",
+      "model.layers.16.linear_attn.in_proj_ba",
+      "model.layers.16.mlp.gate",
+      "model.layers.16.mlp.shared_expert_gate",
+      "model.layers.17.linear_attn.in_proj_ba",
+      "model.layers.17.mlp.gate",
+      "model.layers.17.mlp.shared_expert_gate",
+      "model.layers.18.linear_attn.in_proj_ba",
+      "model.layers.18.mlp.gate",
+      "model.layers.18.mlp.shared_expert_gate",
+      "model.layers.19.mlp.gate",
+      "model.layers.19.mlp.shared_expert_gate",
+      "model.layers.20.linear_attn.in_proj_ba",
+      "model.layers.20.mlp.gate",
+      "model.layers.20.mlp.shared_expert_gate",
+      "model.layers.21.linear_attn.in_proj_ba",
+      "model.layers.21.mlp.gate",
+      "model.layers.21.mlp.shared_expert_gate",
+      "model.layers.22.linear_attn.in_proj_ba",
+      "model.layers.22.mlp.gate",
+      "model.layers.22.mlp.shared_expert_gate",
+      "model.layers.23.mlp.gate",
+      "model.layers.23.mlp.shared_expert_gate",
+      "model.layers.24.linear_attn.in_proj_ba",
+      "model.layers.24.mlp.gate",
+      "model.layers.24.mlp.shared_expert_gate",
+      "model.layers.25.linear_attn.in_proj_ba",
+      "model.layers.25.mlp.gate",
+      "model.layers.25.mlp.shared_expert_gate",
+      "model.layers.26.linear_attn.in_proj_ba",
+      "model.layers.26.mlp.gate",
+      "model.layers.26.mlp.shared_expert_gate",
+      "model.layers.27.mlp.gate",
+      "model.layers.27.mlp.shared_expert_gate",
+      "model.layers.28.linear_attn.in_proj_ba",
+      "model.layers.28.mlp.gate",
+      "model.layers.28.mlp.shared_expert_gate",
+      "model.layers.29.linear_attn.in_proj_ba",
+      "model.layers.29.mlp.gate",
+      "model.layers.29.mlp.shared_expert_gate",
+      "model.layers.30.linear_attn.in_proj_ba",
+      "model.layers.30.mlp.gate",
+      "model.layers.30.mlp.shared_expert_gate",
+      "model.layers.31.mlp.gate",
+      "model.layers.31.mlp.shared_expert_gate",
+      "model.layers.32.linear_attn.in_proj_ba",
+      "model.layers.32.mlp.gate",
+      "model.layers.32.mlp.shared_expert_gate",
+      "model.layers.33.linear_attn.in_proj_ba",
+      "model.layers.33.mlp.gate",
+      "model.layers.33.mlp.shared_expert_gate",
+      "model.layers.34.linear_attn.in_proj_ba",
+      "model.layers.34.mlp.gate",
+      "model.layers.34.mlp.shared_expert_gate",
+      "model.layers.35.mlp.gate",
+      "model.layers.35.mlp.shared_expert_gate",
+      "model.layers.36.linear_attn.in_proj_ba",
+      "model.layers.36.mlp.gate",
+      "model.layers.36.mlp.shared_expert_gate",
+      "model.layers.37.linear_attn.in_proj_ba",
+      "model.layers.37.mlp.gate",
+      "model.layers.37.mlp.shared_expert_gate",
+      "model.layers.38.linear_attn.in_proj_ba",
+      "model.layers.38.mlp.gate",
+      "model.layers.38.mlp.shared_expert_gate",
+      "model.layers.39.mlp.gate",
+      "model.layers.39.mlp.shared_expert_gate",
+      "model.layers.40.linear_attn.in_proj_ba",
+      "model.layers.40.mlp.gate",
+      "model.layers.40.mlp.shared_expert_gate",
+      "model.layers.41.linear_attn.in_proj_ba",
+      "model.layers.41.mlp.gate",
+      "model.layers.41.mlp.shared_expert_gate",
+      "model.layers.42.linear_attn.in_proj_ba",
+      "model.layers.42.mlp.gate",
+      "model.layers.42.mlp.shared_expert_gate",
+      "model.layers.43.mlp.gate",
+      "model.layers.43.mlp.shared_expert_gate",
+      "model.layers.44.linear_attn.in_proj_ba",
+      "model.layers.44.mlp.gate",
+      "model.layers.44.mlp.shared_expert_gate",
+      "model.layers.45.linear_attn.in_proj_ba",
+      "model.layers.45.mlp.gate",
+      "model.layers.45.mlp.shared_expert_gate",
+      "model.layers.46.linear_attn.in_proj_ba",
+      "model.layers.46.mlp.gate",
+      "model.layers.46.mlp.shared_expert_gate",
+      "model.layers.47.mlp.gate",
+      "model.layers.47.mlp.shared_expert_gate",
+      "lm_head"
+    ],
+    "kv_cache_scheme": null,
+    "quant_method": "compressed-tensors",
+    "quantization_status": "compressed",
+    "sparsity_config": {},
+    "transform_config": {},
+    "version": "0.14.0.1"
+  },
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 5000000,
+  "router_aux_loss_coef": 0.001,
+  "shared_expert_intermediate_size": 512,
+  "tie_word_embeddings": false,
+  "transformers_version": "4.57.6",
+  "use_cache": true,
+  "use_sliding_window": false,
+  "vocab_size": 151936
+}

generation_config.json

@@ -0,0 +1,12 @@
+{
+  "bos_token_id": 151643,
+  "do_sample": true,
+  "eos_token_id": [
+    151645,
+    151643
+  ],
+  "pad_token_id": 151643,
+  "top_k": 40,
+  "top_p": 0.95,
+  "transformers_version": "4.57.6"
+}

model-00001-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e34527f9cfa9c35c9083c12970846cbc9a919166cd514bcb318339809c8f81e1
+size 5002386192

model-00002-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3d4dcc33b3937160192dcd07cc039b2f98688649f27935442ab1801bb41d1e43
+size 5003059104

model-00003-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:16289bb7df9f0a51c682bc42b9b5ee07f70f6b2554ce4c2d616648133965daea
+size 5003072496

model-00004-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b0139c08fbf518e4411d8bbba2212ef82a75a4c98d3f4fb6c2d21ac61aa24601
+size 5002916392

model-00005-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fbce2d8dedabb9ebd49b50f6d2ba6b34cbb92e4026ca90bf402e3f00723cc937
+size 5003072496

model-00006-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f933c34a5ca3d1bc63f0a1e20223e4676620682e23947e1a97f82b256fb02ae5
+size 5002855832

model-00007-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:da0396047aaf341e74fa8ef38e7222a5b6b55920054e58b0e1c337d6c6e72bbd
+size 4134020960

model.safetensors.index.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:316d1987a03d301b137bd4af98357ae5aa2824621a2a9d35fad3814aaa7796e2
+size 17323902

recipe.yaml

@@ -0,0 +1,28 @@
+default_stage:
+  default_modifiers:
+    AWQModifier:
+      targets: [Linear]
+      ignore: [lm_head, 're:.*mlp.gate$', 're:.*mlp.shared_expert_gate$', 're:.*linear_attn[.]conv1d',
+        're:.*linear_attn[.]in_proj_a', 're:.*linear_attn[.]in_proj_b']
+      scheme: W4A16
+      bypass_divisibility_checks: false
+      mappings:
+      - smooth_layer: re:.*layers\.(3|7|11|15|19|23|27|31|35|39|43|47)\.input_layernorm$
+        balance_layers: ['re:.*self_attn.q_proj$', 're:.*self_attn.k_proj$', 're:.*self_attn.v_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*self_attn.v_proj$
+        balance_layers: ['re:.*self_attn.o_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*layers\.(0|1|2|4|5|6|8|9|10|12|13|14|16|17|18|20|21|22|24|25|26|28|29|30|32|33|34|36|37|38|40|41|42|44|45|46)\.input_layernorm$
+        balance_layers: ['re:.*linear_attn.in_proj_qkvz$', 're:.*linear_attn.in_proj_ba$']
+        activation_hook_target: null
+      - smooth_layer: re:.*post_attention_layernorm$
+        balance_layers: ['re:.*mlp.experts.*.gate_proj$', 're:.*mlp.experts.*.up_proj$', 're:.*mlp.shared_expert.gate_proj$',
+          're:.*mlp.shared_expert.up_proj$']
+        activation_hook_target: null
+      - smooth_layer: re:.*up_proj$
+        balance_layers: ['re:.*down_proj$']
+        activation_hook_target: null
+      offload_device: !!python/object/apply:torch.device [cpu]
+      duo_scaling: false
+      n_grid: 12

special_tokens_map.json

@@ -0,0 +1,31 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "eos_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aeb13307a71acd8fe81861d94ad54ab689df773318809eed3cbe794b4492dae4
+size 11422654

tokenizer_config.json

@@ -0,0 +1,239 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151665": {
+      "content": "<tool_response>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151666": {
+      "content": "</tool_response>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151667": {
+      "content": "<think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151668": {
+      "content": "</think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": {},
+  "model_max_length": 1048576,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}