github/ColossalAI
1
0
Fork 0
mirror of https://github.com/hpcaitech/ColossalAI.git synced 2025-09-08 12:30:42 +00:00
Code Issues Packages Projects Releases Wiki Activity

[shardformer] support pipeline for deepseek v3 and optimize lora save (#6188)

Browse Source 
* [shardformer] support pipeline for deepseek v3

* [checkpointio] fix lora save

* [devops] update ci env

* [booster] optimize lora

* fix test

* fix test
This commit is contained in:
Hongxin Liu
2025-02-14 14:48:54 +08:00
committed by GitHub
parent ec73f1b5e2
commit 014837e725
21 changed files with 478 additions and 91 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
colossalai/checkpoint_io/checkpoint_io_base.py
View File

@@ -437,7 +437,11 @@ class CheckpointIO(ABC):
@abstractmethod
def save_lora_as_pretrained(
self, model: Union[nn.Module, ModelWrapper], checkpoint: str, use_safetensors: bool = False
self,
model: Union[nn.Module, ModelWrapper],
checkpoint: str,
use_safetensors: bool = False,
state_dict: Optional[dict] = None,
) -> None:
"""
Save the lora adapters and adapter configuration file to a pretrained checkpoint directory.
@@ -446,4 +450,5 @@ class CheckpointIO(ABC):
model (Union[nn.Module, ModelWrapper]): A model boosted by Booster.
checkpoint (str): Path to the checkpoint directory. It must be a local path.
use_safetensors (bool, optional): Whether to use safe tensors when saving. Defaults to False.
state_dict (Optional[dict], optional): The state dict to save. Defaults to None.
"""