423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-10-15 16:18:10 +08:00
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #4580 from hzhaoy/bugfix-deepspeed-pissa

Browse Source 
Fix bug when using pissa method with deepspeed

Former-commit-id: f260d458f91d6d2b4ed141f64844cded11d5aaad
This commit is contained in:
hoshi-hiyouga 2024-06-28 00:46:51 +08:00 committed by GitHub
commit edc7498111
1 changed files with 3 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
src/llamafactory/train/sft/trainer.py
View File

@ -53,6 +53,9 @@ class CustomSeq2SeqTrainer(Seq2SeqTrainer):
self.processor = processor self.processor = processor
if finetuning_args.pissa_convert: if finetuning_args.pissa_convert:
if self.is_deepspeed_enabled:
self.accelerator.deepspeed_config = self.accelerator.state.deepspeed_plugin.deepspeed_config
self.deepspeed = self._wrap_model(self.model_wrapped)
self.save_model(os.path.join(self.args.output_dir, "pissa_init")) self.save_model(os.path.join(self.args.output_dir, "pissa_init"))
if finetuning_args.use_badam: if finetuning_args.use_badam: