biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[AMD][FP8][BugFix] Remove V1 check in arg_utils.py for FP8 since it is not necessary (#17215)

Browse Source 
Signed-off-by: Randall Smith <Randall.Smith@amd.com>
This commit is contained in:
rasmith
2025-04-25 21:55:05 -05:00
committed by GitHub
parent 8de2901fea
commit 68af5f6c5c
2 changed files with 0 additions and 29 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
vllm/engine/arg_utils.py
View File

@@ -1368,23 +1368,6 @@ class EngineArgs:
recommend_to_remove=False)
return False
if current_platform.is_rocm():
from vllm.model_executor.layers.quantization.fp8 import Fp8Config
load_config = self.create_load_config()
quantization_config = VllmConfig.get_quantization_config(
model_config, load_config)
if isinstance(quantization_config, Fp8Config):
_raise_or_fallback(feature_name="fp8 for ROCm",
recommend_to_remove=False)
return False
from vllm.model_executor.layers.quantization.quark.quark import (
QuarkConfig)
if isinstance(quantization_config, QuarkConfig
) and quantization_config.has_fp8_layer_weights():
_raise_or_fallback(feature_name="Quark fp8 for ROCm",
recommend_to_remove=False)
# No Fp8 KV cache so far.
if self.kv_cache_dtype != "auto":
fp8_attention = self.kv_cache_dtype.startswith("fp8")