423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-12-17 20:30:36 +08:00
Code Issues Packages Projects Releases Wiki Activity

[feat] support megatron-LM training by mcore_adapter (#9237)

Browse Source 
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
Co-authored-by: Yaowei Zheng <hiyouga@buaa.edu.cn>
This commit is contained in:
Kingsley
2025-10-26 16:21:30 +08:00
committed by GitHub
parent 129e918106
commit 13170577b2
14 changed files with 671 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
src/llamafactory/hparams/finetuning_args.py
View File

@@ -461,7 +461,7 @@ class FinetuningArguments(
default="sft",
metadata={"help": "Which stage will be performed in training."},
)
finetuning_type: Literal["lora", "freeze", "full"] = field(
finetuning_type: Literal["lora", "oft", "freeze", "full"] = field(
default="lora",
metadata={"help": "Which fine-tuning method to use."},
)
@@ -473,6 +473,10 @@ class FinetuningArguments(
default=False,
metadata={"help": "Whether or not to use the Adam-mini optimizer."},
)
use_mca: bool = field(
default=False,
metadata={"help": "Whether or not to use MCA (Megatron Core Adapter) training. Controlled by USE_MCA environment variable."},
)
use_muon: bool = field(
default=False,
metadata={"help": "Whether or not to use the Muon optimizer."},