[V1] Enable V1 Fp8 cache for FA3 in the oracle (#15191)

Signed-off-by: Lucas Wilkinson <lwilkinson@neuralmagic.com>
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
This commit is contained in:
Lucas Wilkinson
2025-03-23 15:07:04 -07:00
committed by GitHub
parent 9c5c81b0da
commit dccf535f8e
9 changed files with 45 additions and 23 deletions
+1 -1
View File
@@ -205,7 +205,6 @@ from vllm.attention.backends.utils import (PAD_SLOT_ID, compute_slot_mapping,
compute_slot_mapping_start_idx,
is_block_tables_empty)
from vllm.attention.ops.triton_merge_attn_states import merge_attn_states
from vllm.fa_utils import get_flash_attn_version
from vllm.model_executor.layers.linear import (ColumnParallelLinear,
LinearBase, RowParallelLinear,
UnquantizedLinearMethod)
@@ -214,6 +213,7 @@ from vllm.model_executor.layers.rotary_embedding import (
from vllm.multimodal import MultiModalPlaceholderMap
from vllm.platforms import current_platform
from vllm.utils import async_tensor_h2d, cdiv, make_tensor_with_pad, round_down
from vllm.vllm_flash_attn.fa_utils import get_flash_attn_version
try:
from vllm.vllm_flash_attn import flash_attn_varlen_func