apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-07 01:22:59 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[Feature] Set scheduler v1 as default (#3812)

Browse Source 
* [Feature] Set scheduler v1 as default

* [Feature] Set scheduler v1 as default

* [Feature] Set scheduler v1 as default

* [Feature] Set scheduler v1 as default

* [Feature] Set scheduler v1 as default

* [Feature] Set scheduler v1 as default
This commit is contained in:
chenjian
2025-09-04 11:02:10 +08:00
committed by GitHub
parent fbf0e9d2aa
commit fb1e0d6a87
8 changed files with 43 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
fastdeploy/worker/worker_process.py
View File

@@ -740,6 +740,20 @@ 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
fd_config = FDConfig(
model_config=model_config,
parallel_config=parallel_config,