Auto sft #9728
Open
Auto sft #9728
+2,881
−22
Conversation
This file contains 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
blacksheep-Aristotle
force-pushed
the
auto_sft
branch
from
56c9890 to
93041e1
Compare
|
Thanks for your contribution! |
<!-- Demo: PaddlePaddle#26 --> #### Before submitting ```shell # Install and register `pre-commit` in the project folder pip install pre-commit && pre-commit install # Process previous code files separately pre-commit run --file XXXX.py ``` ### PR types <!-- One of [ New features | Bug fixes | Function optimization | Performance optimization | Breaking changes | Others ] --> ### PR changes <!-- One of [ Models | APIs | Docs | Others ] --> ### Description 1.rebase 2.修复sft文件依赖错误，LoRAAutoConfig错误
|
|
<!-- Demo: PaddlePaddle#26 --> #### Before submitting - [ ] Lint code. If there are lint issues, please format the code first. ```shell # Install and register `pre-commit` in the project folder pip install pre-commit && pre-commit install # Process previous code files separately pre-commit run --file XXXX.py ``` - [ ] Add test cases into `tests` folder. If there are codecov issues, please add tests cases first. ### PR types <!-- One of [ New features | Bug fixes | Function optimization | Performance optimization | Breaking changes | Others ] --> ### PR changes <!-- One of [ Models | APIs | Docs | Others ] --> ### Description add ci guard
jeff41404
reviewed
Feb 25, 2025
| --auto_parallel_resume_form_hybrid_parallel true \ | ||
| --sharding_parallel_config "enable_stage1_tensor_fusion" \ | ||
| --tensor_parallel_config "enable_mp_async_allreduce" \ | ||
| --pipeline_parallel_config "enable_sharding_comm_overlap enable_dp_comm_overlap enable_overlap_p2p_comm disable_p2p_cache_shape" \ |
There was a problem hiding this comment.
files llm/run_sft_hand.sh and llm/run_lora_hand.sh need to be added?
jeff41404
reviewed
Feb 25, 2025
Comment on lines
+417
to
+419
| if "labels" in batch.keys(): | ||
| value = batch.pop("labels") | ||
| batch["labels"] = value |
There was a problem hiding this comment.
Why do we need to add this code? it seems that this code does not change batch?
jeff41404
reviewed
Feb 25, 2025
Comment on lines
+476
to
+489
| # #NOTE(zhangwl):this may move to wrap | ||
| # trainable_parameters = [p for p in model.parameters() if not p.stop_gradient] | ||
| # trainer.set_optimizer_grouped_parameters(trainable_parameters) | ||
|
|
||
| # def forward_pre_hook(layer, input): | ||
| # print(f"{layer} forward start") | ||
|
|
||
| # def forward_post_hook(layer, input, output): | ||
| # print(f"{layer} forward done") | ||
|
|
||
| # for layer in trainer.model.sublayers(): | ||
| # layer.register_forward_pre_hook(forward_pre_hook) | ||
| # layer.register_forward_post_hook(forward_post_hook) | ||
| # Train |
jeff41404
reviewed
Feb 25, 2025
| @@ -2769,6 +2769,7 @@ def merge_auto_dist_configs(self, configs): | |||
| },.... | |||
| ] | |||
| """ | |||
| # import pdb;pdb.set_trace() | |||
|
The title and description are too simple, the title needs to express: using auto parallel to validate SFT/LORA of llama model. The description needs to include a comparison of loss and performance of SFT/LORA between manual parallel and auto parallel. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
PR types
New features
PR changes
Others
Description
nlp支持中层api/基础api sft训练。
支持sft 全量微调
支持lora 微调训练