composer
composer copied to clipboard
mosaicml
Supercharge Your Model Training
Bumps [vit-pytorch](https://github.com/lucidrains/vit-pytorch) from 0.35.8 to 0.36.0. Commits 29fbf0a begin extending some of the architectures over to 3d, starting with basic ViT 4b8f5bc add link to Flax translation by @conceptofmind See...
![dependabot[bot] avatar](https://avatars.githubusercontent.com/in/29110?v=4 "dependabot[bot] avatar"){kind=avatar}
Bumps [gitpython](https://github.com/gitpython-developers/GitPython) from 3.1.28 to 3.1.29. Commits 19aaa74 prepare next release be3400b fix version replacement in setup.py (#1500) dc140c0 remove package signing See full diff in compare view [
# What does this PR do? Removes C4 dataset. This is currently broken with `datasets` upgrade. We recommend using streaming datasets anyways, so we're just going to get rid of...
# What does this PR do? Allow checkpoint loading for S3 without having to create a load_object_store # What issue(s) does this change relate to? fixes CO-1144 # Before submitting...
# What does this PR do? This PR enables checkpoint saving to S3 using a URI by doing the following: - [x] parses `save_folder` and if it's URI with `s3://`...
# What does this PR do? A bunch of renaming related to `ObjectStore` and `object`. See the commit names for the specific renames # What issue(s) does this change relate...
# What does this PR do? Adding FSDP docs to composer # Before submitting - [ ] Have you read the [contributor guidelines](https://github.com/mosaicml/composer/blob/dev/CONTRIBUTING.md)? - [ ] Is this change a...
# What does this PR do? Deyahp ADE20k streaming dataset # Before submitting - [x] Have you read the [contributor guidelines](https://github.com/mosaicml/composer/blob/dev/CONTRIBUTING.md)? - [ ] Is this change a documentation change...
[WIP] Will replace Fused LayerNorm, since Fused LayerNorm's speedup comes from running in low precision mode. Equivalent converge performance has been verified on standard NLP models (Bert, GPT). Next commits:...
Closes https://mosaicml.atlassian.net/browse/CO-988 and #1441 . For consistency sake, and to ensure that the user knows the model is being mutated, all applicable functional model surgery methods return # of modified...