[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:
@@ -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
|
||||
|
||||
Reference in New Issue
Block a user