apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-12-24 13:28:13 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[FDConfig]Remove reasoning_parser/guided_decoding_backend/disable_any_whitespace/device_ids in FDConfig (#4362)

Browse Source 
* remove devices id

* fix unittest

* fix ce

---------

Co-authored-by: root <root@yqlcc01-sys-rpm12rzmwjd.yqlcc01.baidu.com>
This commit is contained in:
YuanRisheng
2025-10-17 10:40:59 +08:00
committed by GitHub
parent d1637db86a
commit a37c9416ac
14 changed files with 73 additions and 59 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
fastdeploy/worker/worker_process.py
View File

@@ -39,6 +39,7 @@ from fastdeploy.config import (
ParallelConfig,
PlasAttentionConfig,
SpeculativeConfig,
StructuredOutputsConfig,
)
from fastdeploy.input.ernie4_5_tokenizer import Ernie4_5Tokenizer
from fastdeploy.inter_communicator import EngineWorkerQueue as TaskQueue
@@ -744,6 +745,8 @@ def initialize_fd_config(args, ranks: int = 1, local_rank: int = 0) -> FDConfig:
early_stop_config = EarlyStopConfig(args.early_stop_config)
structured_outputs_config: StructuredOutputsConfig = StructuredOutputsConfig(args=vars(args))
# Note(tangbinhan): used for load_checkpoint
model_config.pretrained_config.tensor_parallel_rank = parallel_config.tensor_parallel_rank
model_config.pretrained_config.tensor_parallel_degree = parallel_config.tensor_parallel_size
@@ -792,7 +795,7 @@ def initialize_fd_config(args, ranks: int = 1, local_rank: int = 0) -> FDConfig:
if not current_platform.is_cuda() and not current_platform.is_xpu():
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":
if structured_outputs_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
@@ -813,6 +816,7 @@ def initialize_fd_config(args, ranks: int = 1, local_rank: int = 0) -> FDConfig:
scheduler_config=scheduler_config,
ips=args.ips,
plas_attention_config=plas_attention_config,
structured_outputs_config=structured_outputs_config,
)
update_fd_config_for_mm(fd_config)
if fd_config.load_config.load_choices == "default_v1" and not v1_loader_support(fd_config):