[BF16]support bf16 for RLHF training on deepspeedChat#733
Merged
tjruwase merged 2 commits intodeepspeedai:masterfrom Sep 22, 2023
ys950902:master
Merged
[BF16]support bf16 for RLHF training on deepspeedChat#733tjruwase merged 2 commits intodeepspeedai:masterfrom ys950902:master
tjruwase merged 2 commits intodeepspeedai:masterfrom
ys950902:master
Conversation
ys950902
requested review from
RezaYazdaniAminabadi,
ShadenSmith,
arashb,
awan-10,
conglongli,
duli2012,
eltonzheng,
jeffra,
minjiaz,
mrwyattii,
samyam,
tjruwase,
xiaoxiawu-microsoft and
yaozhewei
as code owners
Contributor
Author
|
Hi, for this pr, I have verified on my side with the script you give for all tree training steps on deepspeed-chat, it can enable bfloat16 with flag "--bf16" added, if not add this flag is by default fp16, if you have some suggestions or update, please comment here, many thanks. |
tjruwase
reviewed
Sep 22, 2023
tjruwase
reviewed
Sep 22, 2023
applications/DeepSpeed-Chat/training/step1_supervised_finetuning/main.py
Outdated
Show resolved
Hide resolved
Contributor
Author
|
Hi @tjruwase, thanks for your feedback, I have followed your suggestions to update this issue, so that can expand easily when more data_type needed in the future, could you please also take a look on it. Thanks a lot! |
tjruwase
approved these changes
Sep 22, 2023
Contributor
|
@ys950902, thanks so much for this great PR. |
hwchen2017
pushed a commit
that referenced
this pull request
Jun 8, 2025
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Hi, this pr is to enable bf16 on deepspeedChat for RLHF training, the config function(get_train_ds_config/get_eval_ds_config) is fixed to FP16 before, add flag "--bf16" to enable bf16 as a option to use, if not add it, fp16 is by default like before.