minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

[Cherry pick to release/v0.5.15] Fix NVFP4 online quantization (#30397)

Browse Source 
Co-authored-by: Brayden Zhong <brayden@radixark.ai>
This commit is contained in:
Brayden Zhong
2026-07-07 13:20:36 -07:00
committed by GitHub
co-authored by Brayden Zhong
parent d88644b430
commit e2ea7aafad
1 changed files with 1 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/sglang/srt/layers/quantization/nvfp4_online.py
+1
View File
@@ -85,6 +85,7 @@ class NvFp4OnlineConfig(ModelOptQuantConfig):
self.use_per_token_activation = True self.use_per_token_activation = True
self.is_checkpoint_fp8_serialized = is_checkpoint_fp8_serialized self.is_checkpoint_fp8_serialized = is_checkpoint_fp8_serialized
self.is_fp4_experts = False self.is_fp4_experts = False
self.dequant_fp4_to_fp8 = False
self.activation_scheme = activation_scheme self.activation_scheme = activation_scheme
self.weight_block_size = weight_block_size self.weight_block_size = weight_block_size
self.use_mxfp8 = use_mxfp8 self.use_mxfp8 = use_mxfp8