mirror of
https://github.com/PaddlePaddle/FastDeploy.git
synced 2025-12-24 13:28:13 +08:00
This reverts commit 96d2d4877b.
This commit is contained in:
@@ -38,7 +38,6 @@ from fastdeploy.config import (
|
||||
ModelConfig,
|
||||
ParallelConfig,
|
||||
PlasAttentionConfig,
|
||||
RoutingReplayConfig,
|
||||
SpeculativeConfig,
|
||||
StructuredOutputsConfig,
|
||||
)
|
||||
@@ -886,13 +885,6 @@ def parse_args():
|
||||
help="EPLB Configuration.",
|
||||
)
|
||||
|
||||
parser.add_argument(
|
||||
"--routing_replay_config",
|
||||
type=json.loads,
|
||||
default=None,
|
||||
help="Configation of Rollout Routing Replay.",
|
||||
)
|
||||
|
||||
args = parser.parse_args()
|
||||
return args
|
||||
|
||||
@@ -952,7 +944,6 @@ def initialize_fd_config(args, ranks: int = 1, local_rank: int = 0) -> FDConfig:
|
||||
eplb_config = EPLBConfig(args.eplb_config)
|
||||
|
||||
structured_outputs_config: StructuredOutputsConfig = StructuredOutputsConfig(args=vars(args))
|
||||
routing_replay_config = RoutingReplayConfig(args.routing_replay_config)
|
||||
|
||||
# Note(tangbinhan): used for load_checkpoint
|
||||
model_config.pretrained_config.tensor_parallel_rank = parallel_config.tensor_parallel_rank
|
||||
@@ -1012,7 +1003,6 @@ def initialize_fd_config(args, ranks: int = 1, local_rank: int = 0) -> FDConfig:
|
||||
plas_attention_config=plas_attention_config,
|
||||
structured_outputs_config=structured_outputs_config,
|
||||
eplb_config=eplb_config,
|
||||
routing_replay_config=routing_replay_config,
|
||||
)
|
||||
update_fd_config_for_mm(fd_config)
|
||||
if fd_config.load_config.load_choices == "default_v1" and not v1_loader_support(fd_config):
|
||||
|
||||
Reference in New Issue
Block a user