GitLab

* Fix esm lm head test

* make fixup

* part 1

* part 2

* part 3

* fix

* For CANINE

* For ESMFold
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

* initial commit

* First draft that gets outputs without crashing!

* Add all the ported openfold dependencies

* testing

* Restructure config files for ESMFold

* Debugging to find output discrepancies

* Mainly style

* Make model runnable without extra deps

* Remove utils and merge them to the modeling file

* Use correct gelu and remove some debug prints

* More cleanup

* Update esm docs

* Update conversion script to support ESMFold properly

* Port some top-level changes from ESMFold repo

* Expand EsmFold docstrings

* Make attention_mask optional (default to all 1s)

* Add inference test for ESMFold

* Use config and not n kwargs

* Add modeling output class

* Remove einops

* Remove chunking in ESM FFN

* Update tests for ESMFold

* Quality

* REpo consistency

* Remove tree dependency from ESMFold

* make fixup

* Add an error in case my structure map function breaks later

* Remove needless code

* Stop auto-casting the LM to float16 so CPU tests pass

* Stop auto-casting the LM to float16 so CPU tests pass

* Final test updates

* Split test file

* Copyright and quality

* Unpin PyTorch to see built doc

* Fix config file to_dict() method

* Add some docstrings to the output

* Skip TF checkpoint tests for ESM until we reupload those

* make fixup

* More docstrings

* Unpin to get even with main

* Flag example to write
Co-authored-by: Sylvain Gugger <Sylvain.gugger@gmail.com>

* Partial TF port for ESM model

* Add ESM-TF tests

* Add the various imports for TF-ESM

* TF weight conversion almost ready

* Stop ignoring the decoder weights in PT

* Add tests and lots of fixes

* fix-copies

* Fix imports, add model docs

* Add get_vocab() to tokenizer

* Fix vocab links for pretrained files

* Allow multiple inputs with a sep

* Use EOS as SEP token because ESM vocab lacks SEP

* Correctly return special tokens mask from ESM tokenizer

* make fixup

* Stop testing unsupported embedding resizing

* Handle TF bias correctly

* Skip all models with slow tokenizers in the token classification test

* Fixing the batch/unbatcher of pipelines to accomodate the `None` being

passed around.

* Fixing pipeline bug caused by slow tokenizer  being different.

* Update src/transformers/models/esm/modeling_tf_esm.py
Co-authored-by: Joao Gante <joaofranciscocardosogante@gmail.com>

* Update src/transformers/models/esm/modeling_tf_esm.py
Co-authored-by: Joao Gante <joaofranciscocardosogante@gmail.com>

* Update src/transformers/models/esm/modeling_tf_esm.py
Co-authored-by: Joao Gante <joaofranciscocardosogante@gmail.com>

* Update set_input_embeddings and the copyright notices
Co-authored-by: Your Name <you@example.com>
Co-authored-by: Nicolas Patry <patry.nicolas@protonmail.com>
Co-authored-by: Joao Gante <joaofranciscocardosogante@gmail.com>

* Rebase ESM PR and update all file formats

* Fix test relative imports

* Add __init__.py to the test dir

* Disable gradient checkpointing

* Remove references to TFESM... FOR NOW >:|

* Remove completed TODOs from tests

* Convert docstrings to mdx, fix-copies from BERT

* fix-copies for the README and index

* Update ESM's __init__.py to the modern format

* Add to _toctree.yml

* Ensure we correctly copy the pad_token_id from the original ESM model

* Ensure we correctly copy the pad_token_id from the original ESM model

* Tiny grammar nitpicks

* Make the layer norm after embeddings an optional flag

* Make the layer norm after embeddings an optional flag

* Update the conversion script to handle other model classes

* Remove token_type_ids entirely, fix attention_masking and add checks to convert_esm.py

* Break the copied from link from BertModel.forward to remove token_type_ids

* Remove debug array saves

* Begin ESM-2 porting

* Add a hacky workaround for the precision issue in original repo

* Code cleanup

* Remove unused checkpoint conversion code

* Remove unused checkpoint conversion code

* Fix copyright notices

* Get rid of all references to the TF weights conversion

* Remove token_type_ids from the tests

* Fix test code

* Update src/transformers/__init__.py
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

* Update src/transformers/__init__.py
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

* Update README.md
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

* Add credit

* Remove _ args and __ kwargs in rotary embedding

* Assertively remove asserts

* Replace einsum with torch.outer()

* Fix docstring formatting

* Remove assertions in tokenization

* Add paper citation to ESMModel docstring

* Move vocab list to single line

* Remove ESMLayer from init

* Add Facebook copyrights

* Clean up RotaryEmbedding docstring

* Fix docstring formatting

* Fix docstring for config object

* Add explanation for new config methods

* make fix-copies

* Rename all the ESM- classes to Esm-

* Update conversion script to allow pushing to hub

* Update tests to point at my repo for now

* Set config properly for tests

* Remove the gross hack that forced loss of precision in inv_freq and instead copy the data from the model being converted

* make fixup

* Update expected values for slow tests

* make fixup

* Remove EsmForCausalLM for now

* Remove EsmForCausalLM for now

* Fix padding idx test

* Updated README and docs with ESM-1b and ESM-2 separately (#19221)

* Updated README and docs with ESM-1b and ESM-2 separately

* Update READMEs, longer entry with 3 citations

* make fix-copies
Co-authored-by: Your Name <you@example.com>
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
Co-authored-by: Tom Sercu <tsercu@fb.com>
Co-authored-by: Your Name <you@example.com>