Fix torch compile errors (#5158)

2025-04-08 23:04:37 +08:00
parent cc88d98ab8
commit 88d6fd9a11
1 changed files with 2 additions and 2 deletions
--- a/python/sglang/srt/layers/quantization/fp8_utils.py
+++ b/python/sglang/srt/layers/quantization/fp8_utils.py
@@ -243,8 +243,8 @@ def apply_fp8_linear(
        if _is_cuda:
            qinput, x_scale = sglang_per_token_quant_fp8(input_2d)
        else:
-            qinput, x_scale = per_token_group_quant_fp8(
-                input_2d, group_size=input_2d.shape[1]
+            qinput, x_scale = ops.scaled_fp8_quant(
+                input_2d, input_scale, use_per_token_if_dynamic=use_per_token_if_dynamic
            )

    if cutlass_fp8_supported: