This is the MXFP4 PTQ version of huizimao/gpt-oss-120b-uncensored-bf16.
Finetuning is done on Amazon FalseReject train set with 800 samples.
PTQ is done with NVIDIA ModelOpt
Evaluation results obtained on Amazon FalseReject test set with 300 samples.
| Model Variants | False refusal rate |
|---|---|
| gpt-oss-120b original (MXFP4) | 70% |
| LoRA (BF16) | 6% |
| LoRA + PTQ (MXFP4) - this model | 24% |
Code example, documentation, and further QAT checkpoints will be released soon.
model-00001-of-00015.safetensors 4.7 GBmodel-00006-of-00015.safetensors 4.6 GBmodel-00008-of-00015.safetensors 4.6 GBmodel-00010-of-00015.safetensors 4.6 GBmodel-00012-of-00015.safetensors 4.6 GBmodel-00014-of-00015.safetensors 4.6 GBmodel-00002-of-00015.safetensors 4.6 GBmodel-00004-of-00015.safetensors 4.6 GBmodel-00005-of-00015.safetensors 4.1 GBmodel-00007-of-00015.safetensors 4.1 GBmodel-00009-of-00015.safetensors 4.1 GBmodel-00011-of-00015.safetensors 4.1 GBmodel-00013-of-00015.safetensors 4.1 GBmodel-00003-of-00015.safetensors 4.1 GBmodel-00015-of-00015.safetensors 3.5 GB有文件？提交种子。
提交种子