[Bugfix] Fix DeepGEMM warmup when using FlashInferFp8DeepGEMMDynamicBlockScaledKernel (#49467)

Signed-off-by: mgoin <mgoin64@gmail.com>
This commit is contained in:
Michael Goin
2026-07-22 16:28:49 -07:00
committed by GitHub
parent 4b594b4aa1
commit 917fdb5bf7
+14 -4
View File
@@ -133,6 +133,18 @@ def _extract_data_from_fused_moe_module(
return w13, w13_s, w2, w2_s, num_topk
def _is_deep_gemm_backed_kernel(fp8_linear: object) -> bool:
"""
Return True if the selected linear kernel dispatches to DeepGEMM, either
directly or as the fallback branch of a dynamic wrapper.
"""
if isinstance(fp8_linear, DeepGemmFp8BlockScaledMMKernel):
return True
return isinstance(
getattr(fp8_linear, "fallback", None), DeepGemmFp8BlockScaledMMKernel
)
def _fp8_linear_may_use_deep_gemm(module: torch.nn.Module) -> bool:
"""
Return True if the input module/layer could be processed with DeepGEMM.
@@ -147,10 +159,8 @@ def _fp8_linear_may_use_deep_gemm(module: torch.nn.Module) -> bool:
):
return False
if not isinstance(
getattr(module.quant_method, "fp8_linear", None),
DeepGemmFp8BlockScaledMMKernel,
):
fp8_linear = getattr(module.quant_method, "fp8_linear", None)
if not _is_deep_gemm_backed_kernel(fp8_linear):
return False
block_size = get_mk_alignment_for_contiguous_layout()[0]