Separate attention backends (#3005)

This commit is contained in:
Woosuk Kwon
2024-03-07 01:45:50 -08:00
committed by GitHub
parent cbf4c05b15
commit 2daf23ab0c
35 changed files with 561 additions and 271 deletions
+1 -1
View File
@@ -3,7 +3,7 @@ import pytest
import time
import torch
from vllm.model_executor.layers.triton_kernel.prefix_prefill import (
from vllm.model_executor.layers.attention.ops.prefix_prefill import (
context_attention_fwd)
from xformers import ops as xops
from xformers.ops.fmha.attn_bias import BlockDiagonalCausalFromBottomRightMask