Skip to content

GitLab

Commit 2b4f2bc3 authored by zhuwenwen's avatar zhuwenwen
Browse files

update fa backend on gfx926 and gfx928

parent 58117664
Show whitespace changes
Inline Side-by-side
Showing with 1 addition and 1 deletion
vllm/platforms/rocm.py
View file @ 2b4f2bc3
...@@ -154,7 +154,7 @@ def use_rocm_custom_paged_attention( ...@@ -154,7 +154,7 @@ def use_rocm_custom_paged_attention(
sinks: Optional[torch.Tensor] = None) -> bool: sinks: Optional[torch.Tensor] = None) -> bool:
GPU_ARCH = torch.cuda.get_device_properties("cuda").gcnArchName GPU_ARCH = torch.cuda.get_device_properties("cuda").gcnArchName
ON_GFX9 = any(arch in GPU_ARCH for arch in ["gfx90a", "gfx942", "gfx950"]) ON_GFX9 = any(arch in GPU_ARCH for arch in ["gfx90a", "gfx942", "gfx950", "gfx928", "gfx936"])
ON_GFX11_GFX12 = any(arch in GPU_ARCH for arch in ["gfx11", "gfx12"]) ON_GFX11_GFX12 = any(arch in GPU_ARCH for arch in ["gfx11", "gfx12"])
# custom paged attn always supported on V0. On V1, requires sliding window # custom paged attn always supported on V0. On V1, requires sliding window
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment