What's new
Added 🎉
- Added an
Evaluatorclass to make comparing source, target, and predictions easier. - Added a way to resize the vocabulary in the T5 module
- Added an argument
reinit_modulestocached_transformers.get()that allows you to re-initialize the pretrained weights of a transformer model, using layer indices or regex strings. - Added attribute
_should_validate_this_epochtoGradientDescentTrainerthat controls whether validation is run at the end of each epoch. - Added
ShouldValidateCallbackthat can be used to configure the frequency of validation during training. - Added a
MaxPoolingSpanExtractor. ThisSpanExtractorrepresents each span by a component wise max-pooling-operation.
Fixed ✅
- Fixed the docstring information for the
FBetaMultiLabelMeasuremetric. - Various fixes for Python 3.9
- Fixed the name that the
push-to-hfcommand uses to store weights. FBetaMultiLabelMeasurenow works with multiple dimensions- Support for inferior operating systems when making hardlinks
- Use
,as a separator for filenames in theevaluatecommand, thus allowing for URLs (eg.gs://...) as input files. - Removed a spurious error message "'torch.cuda' has no attribute '_check_driver'" that would be appear in the logs
when aConfigurationErrorfor missing GPU was raised. - Load model on CPU post training to save GPU memory.
- Fixed a bug in
ShouldValidateCallbackthat leads to validation occuring after the first epoch regardless ofvalidation_startvalue. - Fixed a bug in
ShouldValidateCallbackthat leads to validation occuring everyvalidation_interval + 1epochs, instead of everyvalidation_intervalepochs. - Fixed a bug in
ShouldValidateCallbackthat leads to validation never occuring at the end of training.
Removed 👋
- Removed Tango components, since they now live at https://github.com/allenai/tango.
- Removed dependency on the
overridespackage
Commits
dd5a010 Evaluator (#5445)
0b54fb0 Bump fairscale from 0.4.4 to 0.4.5 (#5545)
2deacfe Fix should validate callback train end (#5542)
2cdb874 Bump mypy from 0.910 to 0.931 (#5538)
a91946a Keep NLTK down. They broke the download of omw. (#5540)
73a5cfc Removes stuff that now lives in the tango repo (#5482)
1278f16 Move changes from #5534 to correct place. (#5535)
a711703 Fix ShouldValidateCallback (#5536)
b0b3ad4 Update mkdocs-material requirement from <8.1.0,>=5.5.0 to >=5.5.0,<8.2.0 (#5503)
a3d7125 Max out span extractor (#5520)
515fe9b Configure validation frequency (#5534)
d7e0c87 Update transformers requirement from <4.15,>=4.1 to >=4.1,<4.16 (#5528)
4233247 Bump fairscale from 0.4.3 to 0.4.4 (#5525)
71f2d79 fix 'check_for_gpu' (#5522)
06ec7f9 Reinit layers of pretrained transformer in cached_transformers.get() (#5505)
ec1fb69 add missing nltk download in CI (#5529)
ab4f7b5 Fix model loading on GPU post training (#5518)
3552842 Fix moving average args not rendering properly in docs (#5516)
87ad006 Update transformers requirement from <4.13,>=4.1 to >=4.1,<4.15 (#5515)
39f4f4c tick version for nightly releases
38436d8 Use comma as filename separator (#5506)
e0ee7f4 Dimensions in FBetaMultiLabelMeasure (#5501)
d77ba3d Hardlink or copy (#5502)
dbcbcf1 Add installation instructions through conda-forge (#5498)
ebad9ee Bump black from 21.11b1 to 21.12b0 (#5496)
82b1f4f Use the correct filename when uploading models to the HF Hub (#5499)
19f6c8f Resize T5 Vocab (#5497)
c557d51 enforce reading in utf-8 encoding (#5476)
1caf0da Removes dependency on the overrides package (#5490)
b99376f Python 3.9 (#5489)
666eaa5 Update mkdocs-material requirement from <7.4.0,>=5.5.0 to >=5.5.0,<8.1.0 (#5486)
64b2c07 Bump fairscale from 0.4.2 to 0.4.3 (#5474)
0a794c6 Fix metric docstring (#5475)
f86ff9f Bump black from 21.10b0 to 21.11b1 (#5473)
a7f6cdf update cached-path (#5477)
844acfa Update filelock requirement from <3.4,>=3.3 to >=3.3,<3.5 (#5469)
05fc7f6 Bump fairscale from 0.4.0 to 0.4.2 (#5461)
923dbde Bump black from 21.9b0 to 21.10b0 (#5453)
09e22aa Update spacy requirement from <3.2,>=2.1.0 to >=2.1.0,<3.3 (#5460)
54b92ae HF now raises ValueError (#5464)