From 37bd8d6e4c6e37e11ac69cc8844c57ab45dcee3c Mon Sep 17 00:00:00 2001 From: Wentao Ye <44945378+yewentao256@users.noreply.github.com> Date: Sat, 19 Jul 2025 02:25:22 -0400 Subject: [PATCH] [Bug] DeepGemm: Fix TypeError: per_block_cast_to_fp8() missing 1 required positional argument: 'use_ue8m0' for SM100 (#21187) Signed-off-by: yewentao256 --- vllm/utils/deep_gemm.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/vllm/utils/deep_gemm.py b/vllm/utils/deep_gemm.py index 56326c9315bae..8b5713e02c950 100644 --- a/vllm/utils/deep_gemm.py +++ b/vllm/utils/deep_gemm.py @@ -99,7 +99,7 @@ def fp8_m_grouped_gemm_nt_masked(*args, **kwargs): def per_block_cast_to_fp8(x, *args, **kwargs): if _per_block_cast_impl is not None and is_blackwell_deep_gemm_used(): - return _per_block_cast_impl(x) + return _per_block_cast_impl(x, use_ue8m0=True) # TODO: refactor the `per_block_cast_to_fp8` from tests to vllm utils from tests.kernels.quant_utils import per_block_cast_to_fp8 as _pbcf return _pbcf(x, *args, **kwargs)