Add ResNet-9 for CIFAR-10 #193
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
coryMosaicML
pushed a commit
to coryMosaicML/composer
that referenced
this pull request
Feb 23, 2022
This is the architecture [1] that someone got to train to 94% accuracy in 26 seconds [2]. The purpose of adding this is that it should let us do CIFAR-10 runs much faster for quick prototyping. * Adds a composer/models/resnet9_cifar10 directory * Adds a composer/yamls/models/resnet9_cifar10.yaml file * Adds imports / constants where necessary to make the model accessible and not break the unit tests [1] https://myrtle.ai/learn/how-to-train-your-resnet-4-architecture/ [2] https://myrtle.ai/learn/how-to-train-your-resnet-8-bag-of-tricks/
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Outdated suggestions cannot be applied.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 is the architecture that someone got to train to 94% accuracy in 26 seconds. The purpose of adding this is that it should let us do CIFAR-10 runs much faster for quick prototyping--I basically threw this together while waiting for some CIFAR-10 jobs to finish.
Wandb results: 92.82% accuracy from 1 run using our resnet56 hparams. This is slightly lower than in the blog post (94%), but has zero bag-of-tricks enhancements or hparam tuning.