Apply sgl w8a8 fp8 kernel (#3148)

2025-03-09 16:03:32 +08:00
parent 9fb48f951f
commit 0dd6cda288
13 changed files with 523 additions and 37 deletions
--- a/python/sglang/srt/layers/linear.py
+++ b/python/sglang/srt/layers/linear.py
@@ -18,6 +18,7 @@ from sglang.srt.distributed import (
 )
 from sglang.srt.layers.parameter import (
    BasevLLMParameter,
+    BlockQuantScaleParameter,
    PackedColumnParameter,
    PackedvLLMParameter,
    PerTensorScaleParameter,
@@ -27,7 +28,6 @@ from sglang.srt.layers.quantization.base_config import (
    QuantizationConfig,
    QuantizeMethodBase,
 )
-from sglang.srt.layers.quantization.fp8_utils import BlockQuantScaleParameter
 from sglang.srt.utils import set_weight_attrs

 logger = logging.getLogger(__name__)