YuYu1015/Huihui-Gemma-4-E2B-it-abliterated-NVFP4
024
Huihui-gemma-4-E2B-it-abliterated-NVFP4
English
NVFP4 quantization of huihui-ai/Huihui-gemma-4-E2B-it-abliterated, quantized using NVIDIA ModelOpt with NVFP4_MLP_ONLY strategy (only MLP layers quantized, attention preserved in higher precision).
Model Details
Quantization Details
Layers Preserved in Higher Precision
Serving with vLLM
vllm serve /path/to/model \
--quantization modelopt \
--served-model-name gemma-4-e2b \
--trust-remote-code \
--gpu-memory-utilization 0.90 \
--max-model-len 32768 \
--enable-prefix-caching \
--enable-chunked-prefill \
--language-model-onlyDGX Spark (SM121) Compatibility Notes
- NVFP4 on SM121 falls back to W4A16 (native W4A4 path not available, missing
cvt.e2m1x2instruction) - Use
--quantization modelopt(notcompressed-tensors) --language-model-onlyskips vision/audio encoder profiling for text-only inference- Clear page cache before starting on UMA:
sudo sh -c 'echo 3 > /proc/sys/vm/drop_caches'
Safety Warning
This model has safety filtering removed (abliterated) and may generate inappropriate content. Users are solely responsible for all consequences arising from its use.
Credits
- Original Model: google/gemma-4-E2B-it by Google DeepMind
- Abliteration: huihui-ai
- NVFP4 Quantization: YuYu1015 on NVIDIA DGX Spark (GB10)
- Quantization Tool: NVIDIA ModelOpt
繁體中文
huihui-ai/Huihui-gemma-4-E2B-it-abliterated 的 NVFP4 量化版本,使用 NVIDIA ModelOpt 的 NVFP4_MLP_ONLY 策略量化(僅量化 MLP 層,attention 保留高精度)。
模型資訊
量化詳情
保留高精度的層
DGX Spark (SM121) 相容性說明
- NVFP4 在 SM121 上會退回 W4A16(原生 W4A4 路徑不可用,缺少
cvt.e2m1x2指令) - 使用
--quantization modelopt(非compressed-tensors) --language-model-only跳過視覺/音訊編碼器 profiling,加速純文字推理- UMA 架構啟動前請先清除 page cache:
sudo sh -c 'echo 3 > /proc/sys/vm/drop_caches'
安全警告
此模型已移除安全過濾機制(abliterated),可能產生不當內容。使用者須自行承擔所有風險與法律責任。
致謝
- 原始模型:google/gemma-4-E2B-it,Google DeepMind
- 去審查:huihui-ai
- NVFP4 量化:YuYu1015,於 NVIDIA DGX Spark (GB10) 上完成
- 量化工具:NVIDIA ModelOpt
