Delete sgl-kernel AOT bmm_fp8, use flashinfer.bmm_fp8 (#31202)

Co-authored-by: root <root@sgl-b300-inference.datacrunch.io>
2026-07-22 07:44:47 +08:00
parent 1b4cb6b8c1
commit 2f4f2362fb
15 changed files with 63 additions and 237 deletions
@@ -55,7 +55,6 @@ else:
    )
    from sgl_kernel.gemm import (
        awq_dequantize,
-        bmm_fp8,
        dsv3_fused_a_gemm,
        fp8_scaled_mm,
        gptq_gemm,
@@ -153,7 +152,6 @@ else:
        "apply_shuffle_mul_sum",
        "apply_token_bitmask_inplace_cuda",
        "awq_dequantize",
-        "bmm_fp8",
        "build_tree_kernel_efficient",
        "causal_conv1d_fwd",
        "causal_conv1d_update",