diff --git a/examples/llm_ptq/hf_ptq.py b/examples/llm_ptq/hf_ptq.py index 5620ddf6a..bec648bd1 100755 --- a/examples/llm_ptq/hf_ptq.py +++ b/examples/llm_ptq/hf_ptq.py @@ -280,6 +280,7 @@ def auto_quantize( "nvfp4_mlp_only", "nvfp4_experts_only", "nvfp4_omlp_only", + "nvfp4_local_hessian", "mxfp8", ] for qformat in qformat_list diff --git a/examples/llm_ptq/scripts/huggingface_example.sh b/examples/llm_ptq/scripts/huggingface_example.sh index 1b9428834..daf5902b9 100755 --- a/examples/llm_ptq/scripts/huggingface_example.sh +++ b/examples/llm_ptq/scripts/huggingface_example.sh @@ -53,9 +53,9 @@ esac IFS="," for qformat in $QFORMAT; do case $qformat in - fp8 | fp8_pc_pt | fp8_pb_wo | int8_wo | int8_sq | int4_awq | w4a8_awq | fp16 | bf16 | nvfp4 | nvfp4_awq | w4a8_nvfp4_fp8 | w4a8_mxfp4_fp8 | nvfp4_mlp_only | nvfp4_experts_only | nvfp4_omlp_only | nvfp4_svdquant | mxfp8) ;; + fp8 | fp8_pc_pt | fp8_pb_wo | int8_wo | int8_sq | int4_awq | w4a8_awq | fp16 | bf16 | nvfp4 | nvfp4_awq | nvfp4_mse | w4a8_nvfp4_fp8 | w4a8_mxfp4_fp8 | nvfp4_experts_only | nvfp4_mlp_only | nvfp4_omlp_only | nvfp4_svdquant | mxfp8 | nvfp4_local_hessian) ;; *) - echo "Unknown quant argument: Expected one of: [fp8, fp8_pc_pt, fp8_pb_wo, int8_wo, int8_sq, int4_awq, w4a8_awq, fp16, bf16, nvfp4, nvfp4_awq, w4a8_nvfp4_fp8, w4a8_mxfp4_fp8, nvfp4_mlp_only, nvfp4_experts_only, nvfp4_omlp_only, nvfp4_svdquant, mxfp8]" >&2 + echo "Unknown quant argument: Expected one of: [fp8, fp8_pc_pt, fp8_pb_wo, int8_wo, int8_sq, int4_awq, w4a8_awq, fp16, bf16, nvfp4, nvfp4_awq, nvfp4_mse, w4a8_nvfp4_fp8, w4a8_mxfp4_fp8, nvfp4_experts_only, nvfp4_mlp_only, nvfp4_omlp_only, nvfp4_svdquant, mxfp8, nvfp4_local_hessian]" >&2 exit 1 ;; esac