biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Feature] Prefill Context Parallel (PCP) basic support (#28718)

Browse Source 
Signed-off-by: QiuChunshuo <qiuchunshuo@huawei.com>
Signed-off-by: FENP <yuanyongjie.yyj@antgroup.com>
Signed-off-by: LookAround <lixushi@huawei.com>
Signed-off-by: Jingchun Gao <gaojingchun1@huawei.com>
Signed-off-by: zhenwenqi2024 <zhenwenqi_2022@qq.com>
Co-authored-by: FENP <yuanyongjie.yyj@antgroup.com>
Co-authored-by: LookAround <lixushi@huawei.com>
Co-authored-by: Jingchun Gao <gaojingchun1@huawei.com>
Co-authored-by: zhenwenqi2024 <zhenwenqi_2022@qq.com>
Co-authored-by: Jingchun Gao <63247409+gjc0824@users.noreply.github.com>
This commit is contained in:
Qiu
2025-11-20 04:52:44 +08:00
committed by GitHub
parent 02f5903b84
commit 2fd893b4ce
27 changed files with 399 additions and 114 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
vllm/v1/worker/gpu_worker.py
View File

@@ -26,6 +26,7 @@ from vllm.distributed.kv_transfer import (
has_kv_transfer_group,
)
from vllm.distributed.parallel_state import (
get_pcp_group,
get_pp_group,
get_tp_group,
)
@@ -733,6 +734,7 @@ class Worker(WorkerBase):
module.global_num_experts = module.moe_config.num_experts
module.moe_parallel_config = FusedMoEParallelConfig.make(
tp_size_=get_tp_group().world_size,
pcp_size_=get_pcp_group().world_size,
dp_size_=get_dp_group().world_size,
vllm_parallel_config=parallel_config,
)
@@ -886,6 +888,7 @@ def init_worker_distributed_environment(
ensure_model_parallel_initialized(
parallel_config.tensor_parallel_size,
parallel_config.pipeline_parallel_size,
parallel_config.prefill_context_parallel_size,
parallel_config.decode_context_parallel_size,
)