huihui-ai/Huihui-Qwen3.5-35B-A3B-abliterated-NVFP4

This is the NVFP4 quantitative version of huihui-ai/Huihui-Qwen3.5-35B-A3B-abliterated created using vllm-project/llm-compressor

Note

This is just an attempt at NVFP4 quantization; no further tests have been conducted. If there are any issues, please leave a message.

VLLM

1. Environmental installation

uv self update

cd /vllm/vllm-project/
mkdir venv
cd /vllm/vllm-project/venv

uv venv --python 3.12 --seed --managed-python

cd /vllm/vllm-project/venv
source .venv/bin/activate

cd /vllm/vllm-project
git clone ssh://git@github.com/vllm-project/vllm

cd /vllm/vllm-project/vllm
pip install -e .

2. Environment validation

uv pip show vllm
Version: 0.18.1rc1.dev115+gd6bb2a9d9.cu130

uv pip show compressed-tensors
Version: 0.14.0.1

uv pip show transformers
Version: 4.57.6

uv pip show torch
Version: 2.10.0+cu130

uv pip show llmcompressor
warning: Package(s) not found for: llmcompressor

3. Download the model.

cd /vllm/models

hf download huihui-ai/Huihui-Qwen3.5-35B-A3B-abliterated-NVFP4 --local-dir ./huihui-ai/Huihui-Qwen3.5-35B-A3B-abliterated-NVFP4 --token xxx

4. Load the model

cd /vllm/vllm-project/venv
source .venv/bin/activate

export CUDA_VISIBLE_DEVICES=0
unset LD_LIBRARY_PATH

export VLLM_USE_FLASHINFER_MOE_FP4=0
export VLLM_NVFP4_GEMM_BACKEND=cutlass

cd /vllm/models

vllm serve huihui-ai/Huihui-Qwen3.5-35B-A3B-abliterated-NVFP4 \
  --tensor-parallel-size 1 \
  --max-model-len 16384 \
  --trust-remote-code \
  --skip-mm-profiling \
  --limit-mm-per-prompt '{"image": 0, "video": 0}' \
  --enforce-eager \
  --gpu-memory-utilization 0.78 \
  --disable-custom-all-reduce

5. Simple test

curl http://localhost:8000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "huihui-ai/Huihui-Qwen3.5-35B-A3B-abliterated-NVFP4",
    "messages": [
      {"role": "user", "content": "hello"}
    ],
    "temperature": 0.7,
    "top_p": 0.9,
    "max_tokens": 8192,
    "stream": false
  }'

6. Issue

If you encounter the following issue: ValueError: The Tokenizer class TokenizersBackend does not exist or is not currently imported.

Modify tokenizer_config.json

TokenizersBackend-> Qwen2Tokenizer

"tokenizer_class": "TokenizersBackend",

Change it to

"tokenizer_class": "Qwen2Tokenizer",

Usage Warnings

• Risk of Sensitive or Controversial Outputs: This model’s safety filtering has been significantly reduced, potentially generating sensitive, controversial, or inappropriate content. Users should exercise caution and rigorously review generated outputs.

• Not Suitable for All Audiences: Due to limited content filtering, the model’s outputs may be inappropriate for public settings, underage users, or applications requiring high security.

• Legal and Ethical Responsibilities: Users must ensure their usage complies with local laws and ethical standards. Generated content may carry legal or ethical risks, and users are solely responsible for any consequences.

• Research and Experimental Use: It is recommended to use this model for research, testing, or controlled environments, avoiding direct use in production or public-facing commercial applications.

• Monitoring and Review Recommendations: Users are strongly advised to monitor model outputs in real-time and conduct manual reviews when necessary to prevent the dissemination of inappropriate content.

• No Default Safety Guarantees: Unlike standard models, this model has not undergone rigorous safety optimization. huihui.ai bears no responsibility for any consequences arising from its use.

Donation

Your donation helps us continue our further development and improvement, a cup of coffee can do it.

• bitcoin:

  bc1qqnkhuchxw0zqjh2ku3lu4hq45hc6gy84uk70ge

• Support our work on Ko-fi!