[Feature][PCP] Support decode-only FULL CUDA graphs (#53867)

Signed-off-by: QiuChunshuo <qiuchunshuo@huawei.com>
Signed-off-by: QiuChunshuo <chunshuoq@gmail.com>
Co-authored-by: Codex <noreply@openai.com>
Co-authored-by: Lucas Wilkinson <LucasWilkinson@users.noreply.github.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
This commit is contained in:
Qiu Chunshuo
2026-09-15 16:27:46 +08:00
committed by GitHub
co-authored by Codex Lucas Wilkinson mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
parent ca67438c08
commit 676650397f
5 changed files with 283 additions and 42 deletions
+6 -2
View File
@@ -1,5 +1,6 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
from collections.abc import Sequence
from dataclasses import dataclass
from typing import TYPE_CHECKING
@@ -217,9 +218,12 @@ def set_dummy_context(
context_len: int,
num_kv_blocks: int,
max_model_len: int,
input_block_tables: Sequence[torch.Tensor] | None = None,
) -> None:
"""Give each dummy request context_len of context, used when profiling step cost."""
if not block_tables.input_block_tables:
if input_block_tables is None:
input_block_tables = block_tables.input_block_tables
if not input_block_tables:
# Attention-free models have no KV context to fabricate.
return
num_reqs = input_batch.num_reqs
@@ -241,7 +245,7 @@ def set_dummy_context(
seq_len = context_len + query_len
for block_table, block_size, bpk in zip(
block_tables.input_block_tables,
input_block_tables,
block_tables.kernel_block_sizes,
block_tables.blocks_per_kv_block,
):