[ROCm][CI] Fix ROCm LoRA Transformers fallback with full CUDA graphs (#41577)

Signed-off-by: Andreas Karatzas <akaratza@amd.com>
This commit is contained in:
Andreas Karatzas
2026-05-23 04:31:32 +00:00
committed by GitHub
parent 84e351555a
commit d28bdf9344
4 changed files with 88 additions and 19 deletions
+1
View File
@@ -150,6 +150,7 @@ def rocm_unquantized_gemm_impl(
envs.VLLM_ROCM_USE_SKINNY_GEMM
and on_gfx950()
and x.dtype in [torch.float16, torch.bfloat16]
and x.dim() == 2
and (
10 <= n <= 128
and k % 8 == 0