[Attention] Update attention imports (#29540)

Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
This commit is contained in:
Matthew Bonanni
2025-11-27 11:19:09 -05:00
committed by GitHub
parent cd007a53b4
commit fc1d8be3dc
38 changed files with 63 additions and 126 deletions
+1 -2
View File
@@ -8,6 +8,7 @@ import numpy as np
import torch
import torch.nn as nn
from vllm.attention.backends.registry import AttentionBackendEnum
from vllm.config import (
CompilationMode,
CUDAGraphMode,
@@ -157,8 +158,6 @@ class EagleProposer:
)
# Determine allowed attention backends once during initialization.
from vllm.attention.backends.registry import AttentionBackendEnum
self.allowed_attn_types: tuple | None = None
if current_platform.is_rocm():
rocm_types = [TritonAttentionMetadata, FlashAttentionMetadata]