xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-12 08:47:07 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Bugfix] fix qwen3 moe fp8 accuracy issue (#23031)

Browse Source 
Signed-off-by: Jinzhen Lin <jinzhen.ljz@antgroup.com>
This commit is contained in:
Jinzhen Lin 2025-08-17 08:41:23 +08:00 committed by GitHub
parent bf7f470b22
commit a258ad8bcc
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 4 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
vllm/model_executor/layers/quantization/fp8.py
View File

@ -125,6 +125,10 @@ class Fp8Config(QuantizationConfig):
ignored_layers = cls.get_from_keys_or(config, ["ignored_layers"], None) ignored_layers = cls.get_from_keys_or(config, ["ignored_layers"], None)
weight_block_size = cls.get_from_keys_or(config, ["weight_block_size"], weight_block_size = cls.get_from_keys_or(config, ["weight_block_size"],
None) None)
if not ignored_layers:
ignored_layers = cls.get_from_keys_or(config,
["modules_to_not_convert"],
None)
return cls(is_checkpoint_fp8_serialized=is_checkpoint_fp8_serialized, return cls(is_checkpoint_fp8_serialized=is_checkpoint_fp8_serialized,
activation_scheme=activation_scheme, activation_scheme=activation_scheme,
ignored_layers=ignored_layers, ignored_layers=ignored_layers,