GitLab

* initial commit to add decoder caching for T5

* better naming for caching

* finish T5 decoder caching

* correct test

* added extensive past testing for T5

* clean files

* make tests cleaner

* improve docstring

* improve docstring

* better reorder cache

* make style

* Update src/transformers/modeling_t5.py
Co-Authored-By: Yacine Jernite <yjernite@users.noreply.github.com>

* make set output past work for all layers

* improve docstring

* improve docstring
Co-authored-by: Yacine Jernite <yjernite@users.noreply.github.com>

* split beam search and no beam search test

* fix test

* clean generate tests

* add bad words list

* make style

* add bad_words_tokens

* make style

* better naming

* make style

* fix typo

* delete lm_head, skips weight tying
* Fixed s3

* fix conflicts

* update bart max length test

* correct spelling mistakes

* implemented model specific encode function

* fix merge conflicts

* better naming

* save intermediate state -> need to rethink strucuture a bit

* leave tf problem as it is for now

* current version

* add layers.pop

* remove ipdb

* make style

* clean return cut decoding

* remove ipdbs

* Fix restoring layers in the decoders that doesnt exists.

* push good intermediate solution for now

* fix conflicts

* always good to refuse to merge conflicts when rebasing

* fix small bug

* improve function calls

* remove unused file

* add correct scope behavior for t5_generate
Co-authored-by: Morgan Funtowicz <funtowiczmo@gmail.com>

* change do_samples back

* None better default as boolean

* adapt do_sample to True in test example

* make style

* add first copy past test to tf 2 generate

* add tf top_k_top_p_filter fn

* add generate function for TF

* add generate function for TF

* implemented generate for all models expect transfoXL

* implemented generate for all models expect transfoXL

* implemented generate for all models expect transfoXL

* make style

* change permission of test file to correct ones

* delete ipdb

* delete ipdb

* fix bug and finish simple gpt2 integration test

* clean test file

* clean test file

* make style

* make style

* make style

* make style

* change import style

* change import style

* make style

* make style

* add decorators

* add decorators

* fix tf ctrl bug dim => axis in TF

* make style

* make style

* refactored test file

* refactored test file

* take out test_torch_tf_conversion if nothing is defined

* take out test_torch_tf_conversion if nothing is defined

* remove useless files

* remove useless files

* fix conflicts

* fix conflicts

* fix conflicts

* fix conflicts

* fix conflicts

* solve conflicts

* solve conflicts

* fix conflicts

* fix conflicts

* merge conflicts

* delete ipdb

* exposed top_k_top_p_filtering fns

* delete weirdly created w! file

* add comment to test tf common modeling

* fix conflicts

* fix conflicts

* make style

* merge conflicts

* make style

* change tf.tensor.shape to shape_list(tensor)

* correct greedy generation when doing beam search

* improve comment

* Fix tests on GPU (torch)

* Fix bart slow tests
Co-authored-by: Sam Shleifer <sshleifer@gmail.com>

* add slow generate lm_model tests

* fix conflicts

* merge conflicts

* fix conflicts

* add slow generate lm_model tests

* make style

* delete unused variable

* fix conflicts

* fix conflicts

* fix conflicts

* delete unused variable

* fix conflicts

* finished hard coded tests

* improving generation

* finalized special token behaviour for no_beam_search generation

* solved modeling_utils merge conflict

* solve merge conflicts in modeling_utils.py

* add run_generation improvements from PR #2749

* adapted language generation to not use hardcoded -1 if no padding token is available

* remove the -1 removal as hard coded -1`s are not necessary anymore

* add lightweight language generation testing for randomely initialized models - just checking whether no errors are thrown

* add slow language generation tests for pretrained models using hardcoded output with pytorch seed

* delete ipdb

* check that all generated tokens are valid

* renaming

* renaming Generation -> Generate

* make style

* updated so that generate_beam_search has same token behavior than generate_no_beam_search

* consistent return format for run_generation.py

* deleted pretrain lm generate tests -> will be added in another PR

* cleaning of unused if statements and renaming

* run_generate will always return an iterable

* make style

* consistent renaming

* improve naming, make sure generate function always returns the same tensor, add docstring

* add slow tests for all lmhead models

* make style and improve example comments modeling_utils

* better naming and refactoring in modeling_utils

* improving generation

* finalized special token behaviour for no_beam_search generation

* solved modeling_utils merge conflict

* solve merge conflicts in modeling_utils.py

* add run_generation improvements from PR #2749

* adapted language generation to not use hardcoded -1 if no padding token is available

* remove the -1 removal as hard coded -1`s are not necessary anymore

* add lightweight language generation testing for randomely initialized models - just checking whether no errors are thrown

* add slow language generation tests for pretrained models using hardcoded output with pytorch seed

* delete ipdb

* check that all generated tokens are valid

* renaming

* renaming Generation -> Generate

* make style

* updated so that generate_beam_search has same token behavior than generate_no_beam_search

* consistent return format for run_generation.py

* deleted pretrain lm generate tests -> will be added in another PR

* cleaning of unused if statements and renaming

* run_generate will always return an iterable

* make style

* consistent renaming

* improve naming, make sure generate function always returns the same tensor, add docstring

* add slow tests for all lmhead models

* make style and improve example comments modeling_utils

* better naming and refactoring in modeling_utils

* changed fast random lm generation testing design to more general one

* delete in old testing design in gpt2

* correct old variable name

* temporary fix for encoder_decoder lm generation tests - has to be updated when t5 is fixed

* adapted all fast random generate tests to new design

* better warning description in modeling_utils

* better comment

* better comment and error message
Co-authored-by: Thomas Wolf <thomwolf@users.noreply.github.com>

* Results same as fairseq
* Wrote a ton of tests
* Struggled with api signatures
* added some docs

model.parameters() order is apparently not stable (only for xlnet, for some reason)

Created a link between the linear layer bias and the model attribute bias. This does not change anything for the user nor for the conversion scripts, but allows the `resize_token_embeddings` method to resize the bias as well as the weights of the decoder.

Added a test.

It isn't imported anywhere.

I suspect the wrapper classes were created in order to prevent the
abstract base class (TF)CommonModelTester from being included in test
discovery and running, because that would fail.

I solved this by replacing the abstract base class with a mixin.

Code changes are just de-indenting and automatic reformattings
performed by black to use the extra line space.

This construct isn't used anymore these days.

Running python tests/test_foo.py puts the tests/ directory on
PYTHONPATH, which isn't representative of how we run tests.

Use python -m unittest tests/test_foo.py instead.

Do manually what autoflake couldn't manage.

This change is mostly autogenerated with:

    $ python -m autoflake --in-place --recursive examples templates transformers utils hubconf.py setup.py

I made minor changes in the generated diff.

This is the result of:

    $ isort --recursive examples templates transformers utils hubconf.py setup.py

This is the result of:

    $ black --line-length 119 examples templates transformers utils hubconf.py setup.py

There's a lot of fairly long lines in the project. As a consequence, I'm
picking the longest widely accepted line length, 119 characters.

This is also Thomas' preference, because it allows for explicit variable
names, to make the code easier to understand.

Caching models across test cases and across runs of the test suite makes
slow tests somewhat more bearable.

Use gettempdir() instead of /tmp in tests. This makes it easier to
change the location of the cache with semi-standard TMPDIR/TEMP/TMP
environment variables.

Fix #2222.