Skip to content

Commit

Permalink
Use_int8_training and Deepspeed are conflicting
Browse files Browse the repository at this point in the history
  • Loading branch information
xianghuisun authored May 10, 2023
1 parent 97f866c commit d7309fc
Showing 1 changed file with 1 addition and 1 deletion.
2 changes: 1 addition & 1 deletion train/docs/FAQ.md
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,7 @@
这里给出一些实验建议:

1. 不开deepspeed会占用更多显存,建议全量参数finetune模式尽可能采用deepspeed
2. lora训练采用8bit量化训练,不采用deepspeed
2. LoRA训练如果采用8bit量化,就不能使用deepspeed;如果使用deepspeed,就不能指定use_int8_training

关于deepspeed的配置可参考:

Expand Down

0 comments on commit d7309fc

Please sign in to comment.