Release/v1.0 #225
Merged
Release/v1.0 #225
Conversation
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
There was a problem hiding this comment.
Pull request overview
This PR prepares the codebase for v1.0 release by stabilizing training for v9-t models and fixing several bugs. The changes focus on implementing gradient accumulation with warmup, fixing a widespread typo, and enhancing distributed training support.
Key changes include:
- Implementation of a new
GradientAccumulationcallback to dynamically adjust gradient accumulation during training warmup - Updates to the EMA callback to properly account for gradient accumulation steps
- Correction of the "quite" → "quiet" typo throughout configuration and documentation
Reviewed changes
Copilot reviewed 11 out of 11 changed files in this pull request and generated 2 comments.
Show a summary per file
| File | Description |
|---|---|
| yolo/utils/model_utils.py | Added GradientAccumulation callback class for dynamic gradient accumulation during warmup; updated EMA callback to work with gradient accumulation by tracking batch steps |
| yolo/utils/logging_utils.py | Integrated GradientAccumulation into setup; corrected "quite" typo; changed batch progress description; removed clear_live() call |
| yolo/lazy.py | Added configurable accelerator and device settings; enabled sync_batchnorm for distributed training; corrected "quite" typo |
| yolo/config/task/train.yaml | Added equivalent_batch_size parameter to support gradient accumulation configuration |
| yolo/config/general.yaml | Changed device configuration from fixed 0 to auto and added accelerator setting |
| yolo/config/config.py | Added equivalent_batch_size field to DataConfig and accelerator field to Config |
| tests/test_utils/test_bounding_box_utils.py | Removed .to(device) call from model creation in test (model doesn't need to be on device for this test) |
| tests/conftest.py | Removed equivalent_batch_size from test configuration to prevent GradientAccumulation callback activation during tests |
| docs/HOWTO.md | Corrected "quite" → "quiet" in example command |
| README.md | Corrected "quite" → "quiet" in documentation; removed WIP and PWC badges |
| .github/workflows/deploy.yaml | Corrected "quite" → "quiet" in workflow command |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Comment on lines
+79
to
+88
| class GradientAccumulation(Callback): | ||
| def __init__(self, data_cfg: DataConfig, scheduler_cfg: SchedulerConfig): | ||
| super().__init__() | ||
| self.equivalent_batch_size = data_cfg.equivalent_batch_size | ||
| self.actual_batch_size = data_cfg.batch_size | ||
| self.warmup_epochs = getattr(scheduler_cfg.warmup, "epochs", 0) | ||
| self.current_batch = 0 | ||
| self.max_accumulation = 1 | ||
| self.warmup_batches = 0 | ||
| logger.info(":arrows_counterclockwise: Enable Gradient Accumulation") |
There was a problem hiding this comment.
The GradientAccumulation class lacks a docstring explaining its purpose, parameters, and behavior. Consider adding documentation to describe how this callback manages gradient accumulation during training warmup.
Comment on lines
+79
to
+108
| class GradientAccumulation(Callback): | ||
| def __init__(self, data_cfg: DataConfig, scheduler_cfg: SchedulerConfig): | ||
| super().__init__() | ||
| self.equivalent_batch_size = data_cfg.equivalent_batch_size | ||
| self.actual_batch_size = data_cfg.batch_size | ||
| self.warmup_epochs = getattr(scheduler_cfg.warmup, "epochs", 0) | ||
| self.current_batch = 0 | ||
| self.max_accumulation = 1 | ||
| self.warmup_batches = 0 | ||
| logger.info(":arrows_counterclockwise: Enable Gradient Accumulation") | ||
|
|
||
| def setup(self, trainer: "Trainer", pl_module: "LightningModule", stage: str) -> None: | ||
| effective_batch_size = self.actual_batch_size * trainer.world_size | ||
| self.max_accumulation = max(1, round(self.equivalent_batch_size / effective_batch_size)) | ||
| batches_per_epoch = int(len(pl_module.train_loader) / trainer.world_size) | ||
| self.warmup_batches = int(self.warmup_epochs * batches_per_epoch) | ||
|
|
||
| def on_train_epoch_start(self, trainer: "Trainer", pl_module: "LightningModule") -> None: | ||
| self.current_batch = trainer.global_step | ||
|
|
||
| def on_train_batch_start(self, trainer: "Trainer", pl_module: "LightningModule", *args, **kwargs) -> None: | ||
| if self.current_batch < self.warmup_batches: | ||
| current_accumulation = round(lerp(1, self.max_accumulation, self.current_batch, self.warmup_batches)) | ||
| else: | ||
| current_accumulation = self.max_accumulation | ||
| trainer.accumulate_grad_batches = current_accumulation | ||
|
|
||
| def on_train_batch_end(self, trainer: "Trainer", pl_module: "LightningModule", *args, **kwargs) -> None: | ||
| self.current_batch += 1 | ||
|
|
There was a problem hiding this comment.
The new GradientAccumulation callback class lacks test coverage. Consider adding tests to verify the gradient accumulation logic, warmup behavior, and integration with the trainer's accumulate_grad_batches setting.
Member
Author
There was a problem hiding this comment.
@copilot open a new pull request to apply changes based on this feedback
|
@henrytsui000 I've opened a new pull request, #226, to work on those changes. Once the pull request is ready, I'll request review from you. |
11 tasks
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.
Stabilize training for v9-t, and fix some previous bugs.