[QEff. Finetuning] Adding TP+DDP support in HF Trainer stack#907
Draft
smedhe wants to merge 23 commits intoquic:ft_experimental_v1from
Draft
[QEff. Finetuning] Adding TP+DDP support in HF Trainer stack#907smedhe wants to merge 23 commits intoquic:ft_experimental_v1from
smedhe wants to merge 23 commits intoquic:ft_experimental_v1from
Conversation
- Added a logger which will log onto console and file. This code is similar to existing QEff. Finetuning logger code. - Also added dist_utils which serves as utility code when dealing with distributed training. - Added logger test cases for sanity checks. --------- Signed-off-by: Meet Patel <meetkuma@qti.qualcomm.com> Signed-off-by: Sharvari Medhe <smedhe@qti.qualcomm.com> Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
cherry picking PRs- 697,658,667,666,656,652,647,649,645 --------- Signed-off-by: Meet Patel <meetkuma@qti.qualcomm.com> Signed-off-by: Tanisha Chawada <tchawada@qti.qualcomm.com> Signed-off-by: Dhiraj Kumar Sah <dhirajku@qti.qualcomm.com> Signed-off-by: Sharvari Medhe <smedhe@qti.qualcomm.com> Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com> Signed-off-by: Swati Allabadi <sallabad@qti.qualcomm.com> Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
…c#872) we are only cherry-picking PR-787, 791,813,795, skipping rebasing PR 785, cherry-picking experimental related branches from PR 692,747 --------- Signed-off-by: Swati Allabadi <sallabad@qti.qualcomm.com> Signed-off-by: Sharvari Medhe <smedhe@qti.qualcomm.com> Signed-off-by: Tanisha Chawada <tchawada@qti.qualcomm.com> Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com> Co-authored-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com> Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Adding config file to support style remix dataset --------- Signed-off-by: Tanisha Chawada <tchawada@qti.qualcomm.com> Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
…ntation (quic#893) 1) Added unit test cases for Pipeline Parallelism 2) Added documentation on how to run these tests 3) Created a constants file Signed-off-by: Swati Allabadi <sallabad@qti.qualcomm.com> Co-authored-by: Swati Allabadi <sallabad@qti.qualcomm.com> Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Signed-off-by: Ann Kuruvilla <akuruvil@qti.qualcomm.com>
Added testcase to test compare loss and metrics for different sdks to stable sdk Signed-off-by: Tanisha Chawada <tchawada@qti.qualcomm.com>
Updating PP CLI command as per latest changes in config manager In future, this command should also be updated if any changes are done in single SOC CLI command Signed-off-by: Swati Allabadi <sallabad@qti.qualcomm.com> Co-authored-by: Swati Allabadi <sallabad@qti.qualcomm.com>
Signed-off-by: Sharvari Medhe <smedhe@qti.qualcomm.com>
Signed-off-by: Sharvari Medhe <smedhe@qti.qualcomm.com>
Signed-off-by: Sharvari Medhe <smedhe@qti.qualcomm.com>
quic-akuruvil
requested changes
Apr 16, 2026
|
|
||
| git clone https://github.com/quic/efficient-transformers.git | ||
| cd efficient-transformers | ||
| git checkout ft_experimental # Can remove this once merged to mainline |
quic-akuruvil
requested changes
Apr 16, 2026
Contributor
quic-akuruvil
left a comment
quic-akuruvil
left a comment
There was a problem hiding this comment.
Add inference post the checkpoints are saved. And for checkpoint saving stick to native HF checkpoints saving. So it is easier to load model, and add tuned peft adapters.
Signed-off-by: Sharvari Medhe <smedhe@qti.qualcomm.com>
quic-akuruvil
force-pushed
the
ft_experimental_v1
branch
2 times, most recently
from
7258800 to
aff10af
Compare
Signed-off-by: Sharvari Medhe <smedhe@qti.qualcomm.com>
Signed-off-by: Sharvari Medhe <smedhe@qti.qualcomm.com>
quic-akuruvil
force-pushed
the
ft_experimental_v1
branch
2 times, most recently
from
a3a5067 to
69d48ca
Compare
quic-akuruvil
force-pushed
the
ft_experimental_v1
branch
from
13382d8 to
78e96d7
Compare
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
4 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.
Adding TP+DDP support in the new hf trainer stack. It is still an experimental feature due to transformers and accelerate versions mismatch.
Further TO-DO:
load_in_4_bits flag is not supported in v5.1.0, commented out since it is not used
need to add steps for running tp+ddp in the config.md