[Refactor] Remove Fp8OnlineLinearMethod as scheduled (#45463)

Signed-off-by: yewentao256 <zhyanwentao@126.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
This commit is contained in:
Wentao Ye
2026-06-16 04:35:58 -07:00
committed by GitHub
co-authored by mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
parent d53f4593ce
commit 3f53e2138f
5 changed files with 25 additions and 82 deletions
+4 -1
View File
@@ -21,6 +21,9 @@ from vllm.model_executor.layers.quantization.fp8 import (
Fp8LinearMethod,
Fp8MoEMethod,
)
from vllm.model_executor.layers.quantization.online.fp8 import (
Fp8PerTensorOnlineLinearMethod,
)
from vllm.model_executor.model_loader.weight_utils import default_weight_loader
from vllm.platforms import current_platform
@@ -164,7 +167,7 @@ def test_online_quantization(
def check_model(model):
fc1 = model.model.decoder.layers[0].fc1
assert isinstance(fc1.quant_method, Fp8LinearMethod)
assert isinstance(fc1.quant_method, Fp8PerTensorOnlineLinearMethod)
if kv_cache_dtype == "fp8":
attn = model.model.decoder.layers[0].self_attn.attn
assert isinstance(attn.quant_method, Fp8KVCacheMethod)