[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
+5 -4
View File
@@ -777,11 +777,12 @@ def prepare_inputs_to_capture(
input_batch = InputBatch.make_dummy(
num_reqs, num_tokens, input_buffers, max_query_len=max_query_len
)
input_block_tables = block_tables.get_dummy_block_tables(num_reqs)
slot_mapping_provider: BlockTables | PCPManager = block_tables
if pcp_manager is not None:
slot_mapping_provider = pcp_manager
slot_mappings = slot_mapping_provider.get_dummy_slot_mappings(num_tokens)
input_batch = pcp_manager.prepare_inputs_to_capture(input_batch)
block_table_provider = pcp_manager or block_tables
input_block_tables = block_table_provider.get_dummy_block_tables(num_reqs)
slot_mappings = block_table_provider.get_dummy_slot_mappings(num_tokens)
slot_mappings_by_layer = build_slot_mappings_by_layer(
slot_mappings, kv_cache_config
)