apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-05 16:48:03 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

【Sync】Release/2.0.1 (#2745)

Browse Source 
* add rl qwen model support

* fix

* fix
This commit is contained in:
gaoziyuan
2025-07-08 14:38:18 +08:00
committed by GitHub
parent ea787d8f62
commit 6851489425
13 changed files with 438 additions and 171 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
fastdeploy/model_executor/model_loader.py
View File

@@ -91,8 +91,11 @@ class DefaultModelLoader(BaseModelLoader):
def load_model(self, fd_config: FDConfig) -> nn.Layer:
context = paddle.LazyGuard()
architectures = fd_config.model_config.architectures[0]
# TODO(gongshaotian): Now, only support safetensor
model_class = MODEL_CLASSES[architectures]
if fd_config.load_config.dynamic_load_weight:
# register rl model
import fastdeploy.rl
architectures = architectures + "RL"
with context:
model_cls = ModelRegistry.get_class(architectures)
@@ -104,6 +107,8 @@ class DefaultModelLoader(BaseModelLoader):
if fd_config.load_config.dynamic_load_weight:
return model
# TODO(gongshaotian): Now, only support safetensor
model_class = MODEL_CLASSES[architectures]
state_dict = load_composite_checkpoint(
fd_config.parallel_config.model_name_or_path,
model_class,