vllm - 💡(How to fix) Fix [Bug]: v0.22.0 fails to load nvidia/Qwen3.6-35B-A3B-NVFP4: lm_head.input_scale not registered

lm_head.input_scale
lm_head.weight
lm_head.weight_scale
lm_head.weight_scale_2

---

docker run --rm \
  --name vllm-qwen35-nvidia-sci \
  --runtime nvidia \
  --gpus all \
  --ipc=host \
  -p 8082:8000 \
  -v /path/to/qwen3.6-35b-a3b-nvidia-nvfp4:/model:ro \
  vllm/vllm-openai:v0.22.0 \
  /model \
  --served-model-name qwen35-nvidia-nvfp4-sci-thinking \
  --host 0.0.0.0 \
  --port 8000 \
  --api-key local \
  --trust-remote-code \
  --max-model-len 131072 \
  --max-num-seqs 1 \
  --kv-cache-dtype fp8 \
  --generation-config vllm \
  --gpu-memory-utilization 0.94 \
  --max-num-batched-tokens 8192 \
  --default-chat-template-kwargs '{"enable_thinking":true,"preserve_thinking":true}' \
  --speculative-config '{"method":"mtp","num_speculative_tokens":3,"moe_backend":"triton"}' \
  --quantization modelopt \
  --enable-expert-parallel \
  --enable-auto-tool-choice \
  --reasoning-parser qwen3 \
  --tool-call-parser qwen3_xml \
  --enable-chunked-prefill \
  --enable-prefix-caching

---

ValueError: There is no module or parameter named 'lm_head.input_scale' in Qwen3_5MoeForCausalLM.
The available parameters belonging to lm_head (ParallelLMHead) are: {'lm_head.weight'}

---

File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/qwen3_5.py", line 674, in load_weights
  return loader.load_weights(weights, mapper=self.hf_to_vllm_mapper)
...
File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/utils.py", line 337, in _load_module
  raise ValueError(msg)
ValueError: There is no module or parameter named 'lm_head.input_scale' in Qwen3_5MoeForCausalLM. The available parameters belonging to lm_head (ParallelLMHead) are: {'lm_head.weight'}

---

vLLM image: vllm/vllm-openai:v0.22.0
vLLM version: 0.22.0
Image digest: vllm/vllm-openai@sha256:0fec7ec5f3e6bc168e54899935fb0557da908a4832a1dbc88e2debcf2f889416
GPU: NVIDIA GeForce RTX 5090, 32607 MiB
Driver: 610.47
Docker: Docker version 29.5.2, build 79eb04c
OS: WSL2 Linux x86_64