[V0 deprecation] Remove _VLLM_V1 suffixes from attention backend names (#25489)
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com> Signed-off-by: Matthew Bonanni <mbonanni001@gmail.com>
This commit is contained in:
@@ -103,7 +103,7 @@ backend_configs = {
|
||||
# Triton Attention
|
||||
"TritonAttn":
|
||||
BackendConfig(name="TritonAttn",
|
||||
env_vars={"VLLM_ATTENTION_BACKEND": "TRITON_ATTN_VLLM_V1"},
|
||||
env_vars={"VLLM_ATTENTION_BACKEND": "TRITON_ATTN"},
|
||||
comp_config={
|
||||
"cudagraph_mode": "FULL",
|
||||
}),
|
||||
|
||||
@@ -338,7 +338,7 @@ else:
|
||||
@pytest.mark.parametrize("model_name, model_class", MODELS)
|
||||
@pytest.mark.parametrize("backend",
|
||||
[_Backend.FLASHINFER] if current_platform.is_cuda()
|
||||
else [_Backend.TRITON_ATTN_VLLM_V1])
|
||||
else [_Backend.TRITON_ATTN])
|
||||
@pytest.mark.parametrize(
|
||||
"split_attention",
|
||||
[False, True] if current_platform.is_rocm() else [False])
|
||||
|
||||
Reference in New Issue
Block a user