apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-05 08:37:06 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[Feature] Set v1 scheduler as default in develop (#3807)

Browse Source 
* Set scheduler v1 as default

* Set scheduler v1 as default

* Set scheduler v1 as default

* Set scheduler v1 as default

* Set scheduler v1 as default

* close V1 in guided_decoding

* fix vl ci

* close V1 in guided_decoding
This commit is contained in:
chenjian
2025-09-04 15:16:56 +08:00
committed by GitHub
parent e83251699f
commit 22c165d6dd
8 changed files with 50 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
fastdeploy/worker/worker_process.py
View File

@@ -748,6 +748,23 @@ def initialize_fd_config(args, ranks: int = 1, local_rank: int = 0) -> FDConfig:
logger.info(f"- Dynamic load weight: {load_config.dynamic_load_weight}")
logger.info(f"- Load strategy: {load_config.load_strategy}")
if (
args.speculative_config is not None
and ("method" in args.speculative_config)
and (args.speculative_config["method"] is not None)
):
logger.info("Set ENABLE_V1_KVCACHE_SCHEDULER to 0 due to not support speculative decoding now.")
envs.ENABLE_V1_KVCACHE_SCHEDULER = 0
if args.splitwise_role != "mixed":
logger.info(f"Set ENABLE_V1_KVCACHE_SCHEDULER to 0 due to not supported {args.splitwise_role} now.")
envs.ENABLE_V1_KVCACHE_SCHEDULER = 0
if not current_platform.is_cuda():
logger.info("Set ENABLE_V1_KVCACHE_SCHEDULER to 0 due to not supported.")
envs.ENABLE_V1_KVCACHE_SCHEDULER = 0
if parallel_config.guided_decoding_backend != "off":
logger.info("Set ENABLE_V1_KVCACHE_SCHEDULER to 0 due to not supported guided_decoding.")
envs.ENABLE_V1_KVCACHE_SCHEDULER = 0
fd_config = FDConfig(
model_config=model_config,
parallel_config=parallel_config,