[Quant Kernel] refactored per token group quant fp8 to support int8 up-to 2x faster (#4396)

2025-03-23 23:44:17 -07:00
parent 3980ff1be6
commit 65c24c28f9
8 changed files with 191 additions and 127 deletions
--- a/sgl-kernel/python/sgl_kernel/init.py
+++ b/sgl-kernel/python/sgl_kernel/init.py
@@ -31,6 +31,7 @@ from sgl_kernel.gemm import (
    int8_scaled_mm,
    sgl_per_tensor_quant_fp8,
    sgl_per_token_group_quant_fp8,
+    sgl_per_token_group_quant_int8,
    sgl_per_token_quant_fp8,
 )
 from sgl_kernel.moe import moe_align_block_size, topk_softmax