Feature/sg 708 time units #1181
Merged
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
drop_last: True
…to feature/SG-686-torch-compile
…to feature/SG-686-torch-compile # Conflicts: # src/super_gradients/training/utils/callbacks/callbacks.py
…to feature/SG-686-torch-compile
…to feature/SG-686-torch-compile
…to feature/SG-686-torch-compile
LHBuilder
pushed a commit
to LHBuilder/YOLO-NAS
that referenced
this pull request
Jun 25, 2023
* Added torch.compile support * Timer * Timer * Targets fpr torch compile * Disable DDP * train_dataloader_params: drop_last: True * Fix Timer callback to log events per global step * load_backbone: False * Detection models * Lower LR * Added notes * Added per-epoch timers * Fix wrong nesting of drop_last * Fixes to logging * Log values per step/epoch explictly * Fixes to logging * Fixes to logging * Increase num epochs * Update numbers * Added epoch_total_time_sec * cityscapes_stdc_seg50 * load_backbone: False * imagenet_regnetY * imagenet_regnetY * cityscapes_stdc_seg50 with different compilation modes * cityscapes_stdc_seg50 * cityscapes_ddrnet * Update makefile targets * Ensure we log only on master * Add sync point to ensure we've compiled model on all nodes before going further * Reduce bs * Adding makefile targets * Yolo Nas configs * Yolo Nas configs * Add timer * Add timer * segmentation_compile_tests * segmentation_compile_tests * segmentation_compile_tests * Call to torch.compile after we set up DDP * cityscapes_ddrnet_test * Omit to(device) after converting model to syncbn * Change default torch_compile_mode to reduce-overhead * Update makeifle * segmentation_compile_tests * Update makeifle * Filling table * Update makeifle * Filling table * Filling table * Update makeifle * Update makeifle * Update makeifle * Adding time units * Yolo NAS numbers * Yolo NAS numbers * Add import of TimerCallback * Fixed the potential crash if TimerCallback used for evaluate_from_recipe * Fix missing inheritance for GlobalBatchStepNumber
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.
This PR adds the possibility to log scalar values with explicit time units associated with it.
It enables us to log loss value per step, for instance. The motivation of having this feature is three-fold:
torch.compile PR also depends on this PR (We are using TimerCallback to measure training speedup).