* Initial commit
* Another bunch of updates
* make style quliaty + delete debug arg from bash script
* Use compue_metrics func
* Do a few fixes
* Add copyright
* Fix typos
* Add NER example with accelerate library
* This commit contains the first (yet really unfinished)
version of a script for showing how to train HuggingFace model
with their new accelerate library.
* Fix metric calculation
* make style quality
* mv ner_no_trainer to token-classification dir
* Delete --debug flag from running script
* hf_datasets -> raw_datasets
* Make a few slight adjustments
* Add an informative comment + rewrite a help comment
* Change header
* Fix a few things
* Enforce to use fast tokenizers only
* DataCollatorWithPadding -> DataCollatorForTokenClassification
* Change bash script: python3 -> accelerate launch
* make style
* Add a few missing things (see below)
* Add a max-lenghth padding to predictions and labels to
enable accelerate gather functionality
* Add PyTorch no trainer example to the example README.md
* Remove --do-train from args as being redundant for now
* DataCollatorWithPadding -> DataCollatorForTokenClassification
* Remove some obsolete args.do_train conditions from the script
* Delete --do_train from bash running script
* Delete use_slow_tokenizer from args
* Add unintentionally removed flag --label_all_tokens
* Delete --debug flag from running script
* [examples/seq2seq] fix t5 examples
This PR:
* fixes T5 examples to include `--source_prefix` - it's **not** optional. If you give it a try you will see that you get 10x worse bleu scores w/o it. w/ `27.6849`, w/ `2.374`
* added a normal translation example w/o the peculiarities of MBart and T5
* reduces the default max samples to 50 so it's much faster to test quickly
summarization seems to be broken for t5 score-wise: https://github.com/huggingface/transformers/issues/10733
@sgugger
* specify explicitly the t5 models requiring the special handling
* one more
* update the t5 summarization example to use cnn_dailymail
* move max*samples into the top level README.md
* better wording
* better wording
* add initial script
* finish script
* add shell script example
* accept chars_to_ignor as cl arg
* align the script with other example scripts
* add torchaudio dep
* wav2vec2: support datasets other than LibriSpeech
* Formatting run_asr.py to pass code quality test
* bundled orthography options and added verbose logs
* fixing a typo in timit fine-tuning script
* update comment for clarity
* resize_lm_head and load custom vocab from file
* adding a max_duration_in_seconds filter
* do not assign `duration_filter` lambda, use a def
* log untransliterated text as well
* fix base model for arabic
* fix duration filter when target_sr is not set
* drop duration_in_seconds when unneeded
* script for wav2vec2-large-lv60-timit-asr
* fix for "tha" in arabic corpus (huggingface#10581)
* adding more options to work with common_voice
* PR feedback (huggingface#10581)
* small README change
* pass hf optimizer and scheduler to deepspeed if not specified in ds config
* pass hf optimizer and scheduler to deepspeed if not specified in ds config
* update
* make init_deepspeed support config dict
* fix docstring formatting
* clean up trainer's comments
* add new tests
* fix type
* composit argparse doesn't work
* style
* add a new test, rename others
* document new functionality
* complete tests, add docs
* style
* correct level
* Apply suggestions from code review
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
* add new methods to the doc
* must tell DS we are using a non-native optimizer
* add protection against cpu_offload + HF optimizer combo
* fix the cli overrides
* sync docs + tests
* restore AdamW
* better docs
* need new version
* no longer needed
* remove outdate information
* refactor duplicated code
Co-authored-by: Stas Bekman <stas@stason.org>
Co-authored-by: Stas Bekman <stas00@users.noreply.github.com>
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
* offline mode start
* add specific values
* fix fallback
* add test
* better values check and range
* test that actually works
* document the offline mode
* Apply suggestions from code review
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
* more strict check
* cleaner test
* pt-only test
* style
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
* fix run_seq2seq.py; porting DeepSpeed tests to it
* unrefactor
* defensive programming
* defensive programming 2
* port the rest of the trainer tests
* style
* a cleaner scripts dir finder
* cleanup
* remove xnli_compute_metrics, add load_dataset, load_metric, set_seed,metric.compute,load_metric
* fix
* fix
* fix
* push
* fix
* everything works
* fix init
* fix
* special treatment for sepconv1d
* style
* 🙏🏽
* add doc and cleanup
* fix doc
* fix doc again
* fix doc again
* Apply suggestions from code review
* make style
* Proposal that should work
* Remove needless code
* Fix test
* Apply suggestions from code review
* remove xnli_compute_metrics, add load_dataset, load_metric, set_seed,metric.compute,load_metric
* amend README
* removed data_args.task_name and replaced with task_name = "xnli"; use split function to load train and validation dataset separately; remove __post_init__; remove flag --task_name from README.
* removed dict task_to_keys, use str "xnli" instead of variable task_name, change preprocess_function to use examples["premise"], examples["hypothesis"] directly, remove sentence1_key and sentence2_key, change compute_metrics function to cater only to accuracy metric, add condition for train_langauge is None when using dataset.load_dataset()
* removed `torch.distributed.barrier()` and `import torch` as `from_pretrained` is able to do the work; amend README
* change tokenizer requirement
* split line
* Correct typo from list to str
* improve style
* make other function pretty as well
* add comment
* correct typo
* add new test
* pass tests for tok without padding token
* Apply suggestions from code review
* MOD: fit chinese wwm to new datasets
* MOD: move wwm to new folder
* MOD: formate code
* Styling
* MOD add param and recover trainer
Co-authored-by: Sylvain Gugger <sylvain.gugger@gmail.com>
* Pad to 8x for fp16 multiple choice example (#9752)
* Pad to 8x for fp16 squad trainer example (#9752)
* Pad to 8x for fp16 ner example (#9752)
* Pad to 8x for fp16 swag example (#9752)
* Pad to 8x for fp16 qa beam search example (#9752)
* Pad to 8x for fp16 qa example (#9752)
* Pad to 8x for fp16 seq2seq example (#9752)
* Pad to 8x for fp16 glue example (#9752)
* Pad to 8x for fp16 new ner example (#9752)
* update script template #9752
* Update examples/multiple-choice/run_swag.py
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
* Update examples/question-answering/run_qa.py
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
* Update examples/question-answering/run_qa_beam_search.py
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
* improve code quality #9752
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
* Auto-resume training from checkpoint
* Update examples/text-classification/run_glue.py
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>
* Roll out to other examples
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>
* Switch metrics in run_ner to datasets
* Add flag to return all metrics
* Upstream (and rename) sortish_sampler
* Revert "Upstream (and rename) sortish_sampler"
This reverts commit e07d0dcf65.
* Update run_glue for do_predict with local test data (#9442)
* Update run_glue (#9442): fix comments ('files' to 'a file')
* Update run_glue (#9442): reflect the code review
* Update run_glue (#9442): auto format
* Update run_glue (#9442): reflect the code review
* fix a bug in eval_batch_retrieval
* should return parser as well as other staticmethod
* remove duplicate argument
* these kwargs are no longer accepted (cause TypeError in self.generator.generate of modeling_rag.py)
* fixed file paths in README
* moved an arg to add_ray_specific_args
* Add label smoothing in Trainer
* Add options for scheduler and Adafactor in Trainer
* Put Seq2SeqTrainer in the main lib
* Apply suggestions from code review
Co-authored-by: Stas Bekman <stas00@users.noreply.github.com>
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>
* Address review comments and adapt scripts
* Documentation
* Move test not using script to tests folder
Co-authored-by: Stas Bekman <stas00@users.noreply.github.com>
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>
* Update the README of the text classification example
* Update examples/README.md
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>
* Adapt comment from review
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>
* Add new run_swag example
* Add check
* Add sample
* Apply suggestions from code review
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>
* Very important change to make Lysandre happy
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>
* trainer and finetune_trainer enhancements and fixes
* add fallback default
* move the fixing of incorrect keys back into finetune trainer
* s/eval/val/ to match the split
* trainer can now use a different prefix than eval_ for metrics
* document new arg
* Apply suggestions from code review
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
* use 'eval' as the default for metric_key_prefix
* complete adjust var names + disambiguate
* fix logger
* add clarifying comment
* add clarifying comment
* style
* Apply suggestions from code review
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>
* Update src/transformers/trainer.py
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>
* complete removal of optional for metric_key_prefix
* Apply suggestions from code review
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>
* Reorganize example folder
* Continue reorganization
* Change requirements for tests
* Final cleanup
* Finish regroup with tests all passing
* Copyright
* Requirements and readme
* Make a full link for the documentation
* Address review comments
* Apply suggestions from code review
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>
* Add symlink
* Reorg again
* Apply suggestions from code review
Co-authored-by: Thomas Wolf <thomwolf@users.noreply.github.com>
* Adapt title
* Update to new strucutre
* Remove test
* Update READMEs
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>
Co-authored-by: Thomas Wolf <thomwolf@users.noreply.github.com>
* Remove "Model" suffix from Flax models to look more 🤗
Signed-off-by: Morgan Funtowicz <morgan@huggingface.co>
* Initial working (forward + backward) for Flax MLM training example.
Signed-off-by: Morgan Funtowicz <morgan@huggingface.co>
* Simply code
Signed-off-by: Morgan Funtowicz <morgan@huggingface.co>
* Addressing comments, using module and moving to LM task.
Signed-off-by: Morgan Funtowicz <morgan@huggingface.co>
* Restore parameter name "module" wrongly renamed model.
Signed-off-by: Morgan Funtowicz <morgan@huggingface.co>
* Restore correct output ordering...
Signed-off-by: Morgan Funtowicz <morgan@huggingface.co>
* Actually commit the example 😅
Signed-off-by: Morgan Funtowicz <morgan@huggingface.co>
* Add FlaxBertModelForMaskedLM after rebasing.
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>
* Make it possible to initialize the training from scratch
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>
* Reuse flax linen example of cross entropy loss
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>
* Added specific data collator for flax
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>
* Remove todo for data collator
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>
* Added evaluation step
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>
* Added ability to provide dtype to support bfloat16 on TPU
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>
* Enable flax tensorboard output
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>
* Enable jax.pmap support.
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>
* Ensure batches are correctly sized to be dispatched with jax.pmap
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>
* Enable bfloat16 with --fp16 cmdline args
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>
* Correctly export metrics to tensorboard
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>
* Added dropout and ability to use it.
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>
* Effectively enable & disable during training and evaluation steps.
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>
* Oops.
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>
* Enable specifying kernel initializer scale
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>
* Style.
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>
* Added warmup step to the learning rate scheduler.
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>
* Fix typo.
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>
* Print training loss
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>
* Make style
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>
* fix linter issue (flake8)
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>
* Fix model matching
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>
* Fix dummies
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>
* Fix non default dtype on Flax models
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>
* Use the same create_position_ids_from_input_ids for FlaxRoberta
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>
* Make Roberta attention as Bert
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>
* fix copy
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>
* Wording.
Co-authored-by: Marc van Zee <marcvanzee@gmail.com>
Co-authored-by: Marc van Zee <marcvanzee@gmail.com>
* Add new SQUAD example
* Same with a task-specific Trainer
* Address review comment.
* Small fixes
* Initial work for XLNet
* Apply suggestions from code review
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>
* Final clean up and working XLNet script
* Test and debug
* Final working version
* Add new SQUAD example
* Same with a task-specific Trainer
* Address review comment.
* Small fixes
* Initial work for XLNet
* Apply suggestions from code review
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>
* Final clean up and working XLNet script
* Test and debug
* Final working version
* Add tick
* Update README
* Address review comments
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>
Without this fix, training a `BARTForSequenceClassification` model with `run_pl_glue.py` gives `TypeError: forward() got an unexpected keyword argument 'token_type_ids'`, because BART does not have token_type_ids. I've solved this issue in the same way as it's solved for the "distilbert" model, and I can train BART models on SNLI without errors now.
* fix DP case on multi-gpu
* make executable
* test all 3 modes
* use the correct check for distributed
* dp doesn't need a special case
* restore original name
* cleanup
* implement support for run-time dependency version checking
* try not escaping !
* use findall that works on py36
* small tweaks
* autoformatter worship
* simplify
* shorter names
* add support for non-versioned checks
* add deps
* revert
* tokenizers not required, check version only if installed
* make a proper distutils cmd and add make target
* tqdm must be checked before tokenizers
* workaround the DistributionNotFound peculiar setup
* handle the rest of packages in setup.py
* fully sync setup.py's install_requires - to check them all
* nit
* make install_requires more readable
* typo
* Update setup.py
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
* restyle
* add types
* simplify
* simplify2
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
* Support BERT relative position embeddings
* Fix typo in README.md
* Address review comment
* Fix failing tests
* [tiny] Fix style_doc.py check by adding an empty line to configuration_bert.py
* make fix copies
* fix configs of electra and albert and fix longformer
* remove copy statement from longformer
* fix albert
* fix electra
* Add bert variants forward tests for various position embeddings
* [tiny] Fix style for test_modeling_bert.py
* improve docstring
* [tiny] improve docstring and remove unnecessary dependency
* [tiny] Remove unused import
* re-add to ALBERT
* make embeddings work for ALBERT
* add test for albert
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>
* replace init_ddp_connection for index init
* style
* add finetune test
* add test data
* move generate tensors to device
* add test on EM metric
* style
* allow multi process test
* keep gloo process group for retrieval
* add multi-gpu test
* use custom accelerator
* clean test finetune
* minor
* style
* style
* typo
* use python call instead of imported main fumction
* return_dict fix in modeling_rag
* use float32 in retrieval
* store as float32 as well in the custom knowledge dataset example
* style
* rename to finetune_rag
* style
* update readme
* rename utils and callbacks to utils_rag and callbacks_rag
* fix test
* patrick's comments
* generate dummy data in the finetue test script
* remove dummy data files
* style
* <small>tiny typo</small>
* Tokenizers: ability to load from model subfolder
* use subfolder for local files as well
* Uniformize model shortcut name => model id
* from s3 => from huggingface.co
Co-authored-by: Quentin Lhoest <lhoest.q@gmail.com>
* Put models in subfolders
* Styling
* Fix imports in tests
* More fixes in test imports
* Sneaky hidden imports
* Fix imports in doc files
* More sneaky imports
* Finish fixing tests
* Fix examples
* Fix path for copies
* More fixes for examples
* Fix dummy files
* More fixes for example
* More model import fixes
* Is this why you're unhappy GitHub?
* Fix imports in conver command
* Use the CI to identify failing tests
* Remove from all examples and tests
* More default switch
* Fixes
* More test fixes
* More fixes
* Last fixes hopefully
* Use the CI to identify failing tests
* Remove from all examples and tests
* More default switch
* Fixes
* More test fixes
* More fixes
* Last fixes hopefully
* Run on the real suite
* Fix slow tests
* Fixing roberta for slow-fast tests
* WIP getting equivalence on pipelines
* slow-to-fast equivalence - working on question-answering pipeline
* optional FAISS tests
* Pipeline Q&A
* Move pipeline tests to their own test job again
* update tokenizer to add sequence id methods
* update to tokenizers 0.9.4
* set sentencepiecce as optional
* clean up squad
* clean up pipelines to use sequence_ids
* style/quality
* wording
* Switch to use_fast = True by default
* update tests for use_fast at True by default
* fix rag tokenizer test
* removing protobuf from required dependencies
* fix NER test for use_fast = True by default
* fixing example tests (Q&A examples use slow tokenizers for now)
* protobuf in main deps extras["sentencepiece"] and example deps
* fix protobug install test
* try to fix seq2seq by switching to slow tokenizers for now
* Update src/transformers/tokenization_utils_base.py
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>
* Update src/transformers/tokenization_utils_base.py
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>
The new run_ner.py script tries to run prediction on the input
test set `datasets["test"]`, but it should be the tokenized set
`tokenized_datasets["test"]`
* add a multi-gpu job for all example tests
* run only ported tests
* rename
* explain why env is re-activated on each step
* mark all unported/checked tests with @require_torch_non_multigpu_but_fix_me
* style
* Apply suggestions from code review
Co-authored-by: Sam Shleifer <sshleifer@gmail.com>
Co-authored-by: Sam Shleifer <sshleifer@gmail.com>
* add training tests
* correct longformer
* fix docs
* fix some tests
* fix some more train tests
* remove ipdb
* fix multiple edge case model training
* fix funnel and prophetnet
* clean gpt models
* undo renaming of albert
* Add new token classification example
* Remove txt file
* Add test
* With actual testing done
* Less warmup is better
* Update examples/token-classification/run_ner_new.py
Co-authored-by: Thomas Wolf <thomwolf@users.noreply.github.com>
* Address review comments
* Fix test
* Make Lysandre happy
* Last touches and rename
* Rename in tests
* Address review comments
* More run_ner -> run_ner_old
Co-authored-by: Thomas Wolf <thomwolf@users.noreply.github.com>
* use decorator
* remove hardcoded paths
* make the test use more data and do real quality tests
* shave off 10 secs
* add --eval_beams 2, reformat
* reduce train size, use smaller custom dataset
* change TokenClassificationTask class methods to static methods
Since we do not require self in the class methods of TokenClassificationTask we should probably switch to static methods. Also, since the class TokenClassificationTask does not contain a constructor it is currently unusable as is. By switching to static methods this fixes the issue of having to document the intent of the broken class.
Also, since the get_labels and read_examples_from_file methods are ought to be implemented. Static method definitions are unchanged even after inheritance, which means that it can be overridden, similar to other class methods.
* Trigger Build
Co-authored-by: Lysandre <lysandre.debut@reseau.eseo.fr>
* make it possible to invoke testconf.py in both test suites without crashing on having the same option added
* perl -pi -e 's|--make_reports|--make-reports|' to be consistent with other opts
* add `pytest --make-reports` to all CIs (and artifacts)
* fix
* Make line by line optional in run_mlm
* Add option to disable dynamic padding
* Add option to plm too and update README
* Typos
* More typos
* Even more typos
* Apply suggestions from code review
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>
* Finish the cleanup of the language-modeling examples
* Update main README
* Apply suggestions from code review
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>
* Apply suggestions from code review
Co-authored-by: Thomas Wolf <thomwolf@users.noreply.github.com>
* Propagate changes
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>
Co-authored-by: Thomas Wolf <thomwolf@users.noreply.github.com>
* Add a template for example scripts and apply it to mlm
* Formatting
* Fix test
* Add plm script
* Add a template for example scripts and apply it to mlm
* Formatting
* Fix test
* Add plm script
* Add a template for example scripts and apply it to mlm
* Formatting
* Fix test
* Add plm script
* Styling
* move the helper code into testing_utils
* port test_trainer_distributed to work with pytest
* improve docs
* simplify notes
* doc
* doc
* style
* doc
* further improvements
* torch might not be available
* real fix
* Apply suggestions from code review
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
* New run_clm script
* Formatting
* More comments
* Remove unused imports
* Apply suggestions from code review
Co-authored-by: Thomas Wolf <thomwolf@users.noreply.github.com>
* Address review comments
* Change link to the hub
Co-authored-by: Thomas Wolf <thomwolf@users.noreply.github.com>
* better reports
* a whole bunch of reports in their own files
* clean up
* improvements
* github artifacts experiment
* style
* complete the report generator with multiple improvements/fixes
* fix
* save all reports under one dir to easy upload
* can remove temp failing tests
* doc fix
* some cleanup
* Make Seq2Seq Trainer more similar to Trainer
* fix typo
* fix seq2seq trainer
* remove from tests
* remove lock
* remove train files
* delete test files
* correct typo
* check at init
* make sure trainer is not slowed down on TPU
* correct isort
* remove use cache
* fix use cache
* add last use chache = false
Updating the run_squad training script to handle the "longformer" `model_type`. The longformer is trained in the same was as RoBERTa, so I've added the "longformer" `model_type` (that's the right hugginface name for the LongFormer model, right?) everywhere there was a "roberta" `model_type` reference. The longformer (like RoBERTa) doesn't use `token_type_ids` (as I understand from looking at the [longformer notebook](https://github.com/patil-suraj/Notebooks/blob/master/longformer_qa_training.ipynb), which is what gets updated after this change.
This fix might be related to [this issue](https://github.com/huggingface/transformers/issues/7249) with SQuAD training when using run_squad.py
* Start simplification
* More progress
* Finished script
* Address comments and update tests instructions
* Wrong test
* Accept files as inputs and fix test
* Update src/transformers/trainer_utils.py
Co-authored-by: Julien Chaumond <chaumond@gmail.com>
* Fix labels and add combined score
* Add special labels
* Update TPU command
* Revert to old label strategy
* Use model labels
* Fix for STT-B
* Styling
* Apply suggestions from code review
Co-authored-by: Thomas Wolf <thomwolf@users.noreply.github.com>
* Code styling
* Fix review comments
Co-authored-by: Julien Chaumond <chaumond@gmail.com>
Co-authored-by: Thomas Wolf <thomwolf@users.noreply.github.com>
* add CustomHFIndex
* typo in config
* update tests
* add custom dataset example
* clean script
* update test data
* minor in test
* docs
* docs
* style
* fix imports
* allow to pass the indexed dataset directly
* update tests
* use multiset DPR
* address thom and patrick's comments
* style
* update dpr tokenizer
* add output_dir flag in use_own_knowledge_dataset.py
* allow custom datasets in examples/rag/finetune.py
* add test for custom dataset in distributed rag retriever
* splitting fast and slow tokenizers [WIP]
* [WIP] splitting sentencepiece and tokenizers dependencies
* update dummy objects
* add name_or_path to models and tokenizers
* prefix added to file names
* prefix
* styling + quality
* spliting all the tokenizer files - sorting sentencepiece based ones
* update tokenizer version up to 0.9.0
* remove hard dependency on sentencepiece 🎉
* and removed hard dependency on tokenizers 🎉
* update conversion script
* update missing models
* fixing tests
* move test_tokenization_fast to main tokenization tests - fix bugs
* bump up tokenizers
* fix bert_generation
* update ad fix several tokenizers
* keep sentencepiece in deps for now
* fix funnel and deberta tests
* fix fsmt
* fix marian tests
* fix layoutlm
* fix squeezebert and gpt2
* fix T5 tokenization
* fix xlnet tests
* style
* fix mbart
* bump up tokenizers to 0.9.2
* fix model tests
* fix tf models
* fix seq2seq examples
* fix tests without sentencepiece
* fix slow => fast conversion without sentencepiece
* update auto and bert generation tests
* fix mbart tests
* fix auto and common test without tokenizers
* fix tests without tokenizers
* clean up tests lighten up when tokenizers + sentencepiece are both off
* style quality and tests fixing
* add sentencepiece to doc/examples reqs
* leave sentencepiece on for now
* style quality split hebert and fix pegasus
* WIP Herbert fast
* add sample_text_no_unicode and fix hebert tokenization
* skip FSMT example test for now
* fix style
* fix fsmt in example tests
* update following Lysandre and Sylvain's comments
* Update src/transformers/testing_utils.py
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
* Update src/transformers/testing_utils.py
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
* Update src/transformers/tokenization_utils_base.py
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
* Update src/transformers/tokenization_utils_base.py
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
* Initial callback proposal
* Finish various callbacks
* Post-rebase conflicts
* Fix tests
* Don't use something that's not set
* Documentation
* Remove unwanted print.
* Document all models can work
* Add tests + small fixes
* Update docs/source/internal/trainer_utils.rst
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>
* Address review comments
* Fix TF tests
* Real fix this time
* This one should work
* Fix typo
* Really fix typo
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>
* Changed name to all no_... arguments and all references to them, inverting the boolean condition
* Change benchmark tests to use new Benchmark Args
* Update src/transformers/benchmark/benchmark_args_utils.py
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>
* Update src/transformers/benchmark/benchmark.py
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>
* Fix Style. Add --no options in help
* fix some part of tests
* Update src/transformers/benchmark/benchmark_args_utils.py
* Update src/transformers/benchmark/benchmark_args_utils.py
* Update src/transformers/benchmark/benchmark_args_utils.py
* fix all tests
* make style
* add backwards compability
* make backwards compatible
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>
Co-authored-by: fmcurti <fcurti@DESKTOP-RRQURBM.localdomain>
* added rag WIP
* path fix
* Formatting / renaming prior to actual work
* added rag WIP
* path fix
* Formatting / renaming prior to actual work
* added rag WIP
* path fix
* Formatting / renaming prior to actual work
* added rag WIP
* Formatting / renaming prior to actual work
* First commit
* improve comments
* Retrieval evaluation scripts
* refactor to include modeling outputs + MPI retriever
* Fix rag-token model + refactor
* Various fixes + finetuning logic
* use_bos fix
* Retrieval refactor
* Finetuning refactoring and cleanup
* Add documentation and cleanup
* Remove set_up_rag_env.sh file
* Fix retrieval wit HF index
* Fix import errors
* Fix quality errors
* Refactor as per suggestions in https://github.com/huggingface/transformers/pull/6813#issuecomment-687208867
* fix quality
* Fix RAG Sequence generation
* minor cleanup plus initial tests
* fix test
* fix tests 2
* Comments fix
* post-merge fixes
* Improve readme + post-rebase refactor
* Extra dependencied for tests
* Fix tests
* Fix tests 2
* Refactor test requirements
* Fix tests 3
* Post-rebase refactor
* rename nlp->datasets
* RAG integration tests
* add tokenizer to slow integration test and allow retriever to run on cpu
* add tests; fix position ids warning
* change structure
* change structure
* add from encoder generator
* save working solution
* make all integration tests pass
* add RagTokenizer.save/from_pretrained and RagRetriever.save/from_pretrained
* don't save paths
* delete unnecessary imports
* pass config to AutoTokenizer.from_pretrained for Rag tokenizers
* init wiki_dpr only once
* hardcode legacy index and passages paths (todo: add the right urls)
* finalize config
* finalize retriver api and config api
* LegacyIndex index download refactor
* add dpr to autotokenizer
* make from pretrained more flexible
* fix ragfortokengeneration
* small name changes in tokenizer
* add labels to models
* change default index name
* add retrieval tests
* finish token generate
* align test with previous version and make all tests pass
* add tests
* finalize tests
* implement thoms suggestions
* add first version of test
* make first tests work
* make retriever platform agnostic
* naming
* style
* add legacy index URL
* docstrings + simple retrieval test for distributed
* clean model api
* add doc_ids to retriever's outputs
* fix retrieval tests
* finish model outputs
* finalize model api
* fix generate problem for rag
* fix generate for other modles
* fix some tests
* save intermediate
* set generate to default
* big refactor generate
* delete rag_api
* correct pip faiss install
* fix auto tokenization test
* fix faiss install
* fix test
* move the distributed logic to examples
* model page
* docs
* finish tests
* fix dependencies
* fix import in __init__
* Refactor eval_rag and finetune scripts
* start docstring
* add psutil to test
* fix tf test
* move require torch to top
* fix retrieval test
* align naming
* finish automodel
* fix repo consistency
* test ragtokenizer save/load
* add rag model output docs
* fix ragtokenizer save/load from pretrained
* fix tokenizer dir
* remove torch in retrieval
* fix docs
* fixe finetune scripts
* finish model docs
* finish docs
* remove auto model for now
* add require torch
* remove solved todos
* integrate sylvains suggestions
* sams comments
* correct mistake on purpose
* improve README
* Add generation test cases
* fix rag token
* clean token generate
* fix test
* add note to test
* fix attention mask
* add t5 test for rag
* Fix handling prefix in finetune.py
* don't overwrite index_name
Co-authored-by: Patrick Lewis <plewis@fb.com>
Co-authored-by: Aleksandra Piktus <piktus@devfair0141.h2.fair>
Co-authored-by: Aleksandra Piktus <piktus@learnfair5102.h2.fair>
Co-authored-by: Aleksandra Piktus <piktus@learnfair5067.h2.fair>
Co-authored-by: Your Name <you@example.com>
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>
Co-authored-by: Quentin Lhoest <lhoest.q@gmail.com>
* ready for PR
* cleanup
* correct FSMT_PRETRAINED_MODEL_ARCHIVE_LIST
* fix
* perfectionism
* revert change from another PR
* odd, already committed this one
* non-interactive upload workaround
* backup the failed experiment
* store langs in config
* workaround for localizing model path
* doc clean up as in https://github.com/huggingface/transformers/pull/6956
* style
* back out debug mode
* document: run_eval.py --num_beams 10
* remove unneeded constant
* typo
* re-use bart's Attention
* re-use EncoderLayer, DecoderLayer from bart
* refactor
* send to cuda and fp16
* cleanup
* revert (moved to another PR)
* better error message
* document run_eval --num_beams
* solve the problem of tokenizer finding the right files when model is local
* polish, remove hardcoded config
* add a note that the file is autogenerated to avoid losing changes
* prep for org change, remove unneeded code
* switch to model4.pt, update scores
* s/python/bash/
* missing init (but doesn't impact the finetuned model)
* cleanup
* major refactor (reuse-bart)
* new model, new expected weights
* cleanup
* cleanup
* full link
* fix model type
* merge porting notes
* style
* cleanup
* have to create a DecoderConfig object to handle vocab_size properly
* doc fix
* add note (not a public class)
* parametrize
* - add bleu scores integration tests
* skip test if sacrebleu is not installed
* cache heavy models/tokenizers
* some tweaks
* remove tokens that aren't used
* more purging
* simplify code
* switch to using decoder_start_token_id
* add doc
* Revert "major refactor (reuse-bart)"
This reverts commit 226dad15ca.
* decouple from bart
* remove unused code #1
* remove unused code #2
* remove unused code #3
* update instructions
* clean up
* move bleu eval to examples
* check import only once
* move data+gen script into files
* reuse via import
* take less space
* add prepare_seq2seq_batch (auto-tested)
* cleanup
* recode test to use json instead of yaml
* ignore keys not needed
* use the new -y in transformers-cli upload -y
* [xlm tok] config dict: fix str into int to match definition (#7034)
* [s2s] --eval_max_generate_length (#7018)
* Fix CI with change of name of nlp (#7054)
* nlp -> datasets
* More nlp -> datasets
* Woopsie
* More nlp -> datasets
* One last
* extending to support allen_nlp wmt models
- allow a specific checkpoint file to be passed
- more arg settings
- scripts for allen_nlp models
* sync with changes
* s/fsmt-wmt/wmt/ in model names
* s/fsmt-wmt/wmt/ in model names (p2)
* s/fsmt-wmt/wmt/ in model names (p3)
* switch to a better checkpoint
* typo
* make non-optional args such - adjust tests where possible or skip when there is no other choice
* consistency
* style
* adjust header
* cards moved (model rename)
* use best custom hparams
* update info
* remove old cards
* cleanup
* s/stas/facebook/
* update scores
* s/allen_nlp/allenai/
* url maps aren't needed
* typo
* move all the doc / build /eval generators to their own scripts
* cleanup
* Apply suggestions from code review
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>
* Apply suggestions from code review
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>
* fix indent
* duplicated line
* style
* use the correct add_start_docstrings
* oops
* resizing can't be done with the core approach, due to 2 dicts
* check that the arg is a list
* style
* style
Co-authored-by: Sam Shleifer <sshleifer@gmail.com>
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>
* Removed 'tgt_len' and 'ext_len' from Transfomer-XL
* Some changes are still to be done
* Removed 'tgt_len' and 'ext_len' from Transfomer-XL (2)
* Removed comments
* Fixed quality
* Changed warning to info
* adding demo
* Update examples/lxmert/requirements.txt
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>
* Update examples/lxmert/checkpoint.sh
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>
* added user input for .py demo
* updated model loading, data extrtaction, checkpoints, and lots of other automation
* adding normalizing for bounding boxes
* Update requirements.txt
* some optimizations for extracting data
* added data extracting file
* added data extraction file
* minor fixes to reqs and readme
* Style
* remove options
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>
Co-authored-by: Lysandre <lysandre.debut@reseau.eseo.fr>
* Add cache_dir to save features TextDataset
This is in case the dataset is in a RO filesystem, for which is the case
in tests (GKE TPU tests).
* style
* Allow tests in examples to use cuda or fp16,if they are available
The tests in examples didn't use the cuda or fp16 even if they where available.
- The text classification example (`run_glue.py`) didn't use the fp16 even if it was available but
the device was take based on the availablity(cuda/cpu).
- The language-modeling example (`run_language_modeling.py`) was having `--no_cuda` argument
which made the test to work without cuda. This example is having issue when running with fp16
thus it not enabled (got an assertion error for perplexity due to it higher value).
- The cuda and fp16 is not enabled for question-answering example (`run_squad.py`) as it is having a
difference in the f1 score.
- The text-generation example (`run_generation.py`) will take the cuda or fp16 whenever it is available.
Resolves some of: #5057
* Unwanted import of is_apex_available was removed
* Made changes to test examples file to have the pass --fp16 only if cuda and apex is avaliable
- run_glue.py: Removed the check for cuda and fp16.
- run_generation.py: Removed the check for cuda and fp16 also removed unwanted flag creation.
* Incorrectly sorted imports fixed
* The model needs to be converted to half precision
* Formatted single line if condition statement to multiline
* The torch_device also needed to be checked before running the test on examples
- The tests in examples which uses cuda should also depend from the USE_CUDA flag,
similarly to the rest of the test suite. Even if we decide to set USE_CUDA to
True by default, setting USE_CUDA to False should result in the examples not using CUDA
* Format some of the code in test_examples file
* The improper import of is_apex_available was sorted
* Formatted the code to keep the style standards
* The comma at the end of list giving a flake8 issue was fixed
* Import sort was fixed
* Removed the clean_test_dir function as its not used right now
* [testing] switch to a new TestCasePlus + get_auto_remove_tmp_dir() for auto-removal of tmp dirs
* respect after=True for tempfile, simplify code
* comments
* comment fix
* put `before` last in args, so can make debug even faster
* Add more token classification examples
* POS tagging example
* Phrase chunking example
* PR review fixes
* Add conllu to third party list (used in token classification examples)
* replace capsys with the more refined CaptureStderr/CaptureStdout
* Update examples/seq2seq/test_seq2seq_examples.py
Co-authored-by: Sam Shleifer <sshleifer@gmail.com>
* [wip] add get_polynomial_decay_schedule_with_warmup
* style
* add assert
* change lr_end to a much smaller default number
* check for exact equality
* [model_cards] electra-base-turkish-cased-ner (#6350)
* for electra-base-turkish-cased-ner
* Add metadata
Co-authored-by: Julien Chaumond <chaumond@gmail.com>
* Temporarily de-activate TPU CI
* Update modeling_tf_utils.py (#6372)
fix typo: ckeckpoint->checkpoint
* the test now works again (#6371)
* correct pl link in readme (#6364)
* refactor almost identical tests (#6339)
* refactor almost identical tests
* important to add a clear assert error message
* make the assert error even more descriptive than the original bt
* Small docfile fixes (#6328)
* Patch models (#6326)
* TFAlbertFor{TokenClassification, MultipleChoice}
* Patch models
* BERT and TF BERT info
s
* Update check_repo
* Ci GitHub caching (#6382)
* Cache Github Actions CI
* Remove useless file
* Colab button (#6389)
* Add colab button
* Add colab link for tutorials
* Fix links for open in colab (#6391)
* Update src/transformers/optimization.py
consistently use lr_end=1e-7 default
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
* [wip] add get_polynomial_decay_schedule_with_warmup
* style
* add assert
* change lr_end to a much smaller default number
* check for exact equality
* Update src/transformers/optimization.py
consistently use lr_end=1e-7 default
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
* remove dup (leftover from merge)
* convert the test into the new refactored format
* stick to using the current_step as is, without ++
Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com>
Co-authored-by: Julien Chaumond <chaumond@gmail.com>
Co-authored-by: Lysandre <lysandre.debut@reseau.eseo.fr>
Co-authored-by: Alexander Measure <ameasure@gmail.com>
Co-authored-by: Rohit Gupta <rohitgr1998@gmail.com>
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>
* add pl_glue example test
* for now just test that it runs, next validate results of eval or predict?
* complete the run_pl_glue test to validate the actual outcome
* worked on my machine, CI gets less accuracy - trying higher epochs
* match run_pl.sh hparms
* more epochs?
* trying higher lr
* for now just test that the script runs to a completion
* correct the comment
* if cuda is available, add --fp16 --gpus=1 to cover more bases
* style
as discussed with @sshleifer, removing this TODO to switch to a tiny model, since it won't be able to test the results of the evaluation (i.e. the results are meaningless).
* Support for Comet.ml
* Need to import comet first
* Log this model, not the one in the backprop step
* Log args as hyperparameters; use framework to allow fine control
* Log hyperparameters with context
* Apply black formatting
* isort fix integrations
* isort fix __init__
* Update src/transformers/trainer.py
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
* Update src/transformers/trainer.py
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
* Update src/transformers/trainer_tf.py
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
* Address review comments
* Style + Quality, remove Tensorboard import test
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
Co-authored-by: Lysandre <lysandre.debut@reseau.eseo.fr>
* support --lr_scheduler with multiple possibilities
* correct the error message
* add a note about supported schedulers
* cleanup
* cleanup2
* needs the argument default
* style
* add another assert in the test
* implement requested changes
* cleanups
* fix relative import
* cleanup
* enable easy checkout switch
allow having multiple repository checkouts and not needing to remember to rerun 'pip install -e .[dev]' when switching between checkouts and running tests.
* make isort happy
* examples needs one too
* Switch from return_tuple to return_dict
* Fix test
* [WIP] Test TF Flaubert + Add {XLM, Flaubert}{TokenClassification, MultipleC… (#5614)
* Test TF Flaubert + Add {XLM, Flaubert}{TokenClassification, MultipleChoice} models and tests
* AutoModels
Tiny tweaks
* Style
* Final changes before merge
* Re-order for simpler review
* Final fixes
* Addressing @sgugger's comments
* Test MultipleChoice
* Rework TF trainer (#6038)
* Fully rework training/prediction loops
* fix method name
* Fix variable name
* Fix property name
* Fix scope
* Fix method name
* Fix tuple index
* Fix tuple index
* Fix indentation
* Fix variable name
* fix eval before log
* Add drop remainder for test dataset
* Fix step number + fix logging datetime
* fix eval loss value
* use global step instead of step + fix logging at step 0
* Fix logging datetime
* Fix global_step usage
* Fix breaking loop + logging datetime
* Fix step in prediction loop
* Fix step breaking
* Fix train/test loops
* Force TF at least 2.2 for the trainer
* Use assert_cardinality to facilitate the dataset size computation
* Log steps per epoch
* Make tfds compliant with TPU
* Make tfds compliant with TPU
* Use TF dataset enumerate instead of the Python one
* revert previous commit
* Fix data_dir
* Apply style
* rebase on master
* Address Sylvain's comments
* Address Sylvain's and Lysandre comments
* Trigger CI
* Remove unused import
* Switch from return_tuple to return_dict
* Fix test
* Add recent model
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>
Co-authored-by: Julien Plu <plu.julien@gmail.com>
* Fully rework training/prediction loops
* fix method name
* Fix variable name
* Fix property name
* Fix scope
* Fix method name
* Fix tuple index
* Fix tuple index
* Fix indentation
* Fix variable name
* fix eval before log
* Add drop remainder for test dataset
* Fix step number + fix logging datetime
* fix eval loss value
* use global step instead of step + fix logging at step 0
* Fix logging datetime
* Fix global_step usage
* Fix breaking loop + logging datetime
* Fix step in prediction loop
* Fix step breaking
* Fix train/test loops
* Force TF at least 2.2 for the trainer
* Use assert_cardinality to facilitate the dataset size computation
* Log steps per epoch
* Make tfds compliant with TPU
* Make tfds compliant with TPU
* Use TF dataset enumerate instead of the Python one
* revert previous commit
* Fix data_dir
* Apply style
* rebase on master
* Address Sylvain's comments
* Address Sylvain's and Lysandre comments
* Trigger CI
* Remove unused import
* DataParallel fixes:
1. switched to a more precise check
- if self.args.n_gpu > 1:
+ if isinstance(model, nn.DataParallel):
2. fix tests - require the same fixup under DataParallel as the training module
* another fix
* Add deebert code
* Add readme of deebert
* Add test for deebert
Update test for Deebert
* Update DeeBert (README, class names, function refactoring); remove requirements.txt
* Format update
* Update test
* Update readme and model init methods
* Added data collator for XLNet language modeling and related calls
Added DataCollatorForXLNetLanguageModeling in data/data_collator.py
to generate necessary inputs for language modeling training with
XLNetLMHeadModel. Also added related arguments, logic and calls in
examples/language-modeling/run_language_modeling.py.
Resolves: #4739, #2008 (partially)
* Changed name to `DataCollatorForPermutationLanguageModeling`
Changed the name of `DataCollatorForXLNetLanguageModeling` to the more general `DataCollatorForPermutationLanguageModelling`.
Removed the `--mlm` flag requirement for the new collator and defined a separate `--plm_probability` flag for its use.
CTRL uses a CLM loss just like GPT and GPT-2, so should work out of the box with this script (provided `past` is taken care of
similar to `mems` for XLNet).
Changed calls and imports appropriately.
* Added detailed comments, changed variable names
Added more detailed comments to `DataCollatorForPermutationLanguageModeling` in `data/data_collator.py` to explain working. Also cleaned up variable names and made them more informative.
* Added tests for new data collator
Added tests in `tests/test_trainer.py` for DataCollatorForPermutationLanguageModeling based on those in DataCollatorForLanguageModeling. A specific test has been added to check for odd-length sequences.
* Fixed styling issues
Otherwise, if label is not specified, the following error occurs:
Traceback (most recent call last):
File "run_ner.py", line 303, in <module>
main()
File "run_ner.py", line 101, in main
model_args, data_args, training_args = parser.parse_json_file(json_file=os.path.abspath(sys.argv[1]))
File "/home/user/anaconda3/envs/bert/lib/python3.7/site-packages/transformers/hf_argparser.py", line 159, in parse_json_file
obj = dtype(**inputs)
TypeError: __init__() missing 1 required positional argument: 'labels'
* Fix the bug 'Attempted relative import with no known parent package' when using the bertabs example. Also change the used model from bertabs-finetuned-cnndm, since it seems not be accessible anymore
* Update run_summarization.py
Co-authored-by: Kevin Canwen Xu <canwenxu@126.com>
* remove references to old API in docstring - update data processors
* style
* fix tests - better type checking error messages
* better type checking
* include awesome fix by @LysandreJik for #5310
* updated doc and examples
* Add BERT Loses Patience (Patience-based Early Exit)
* update model archive
* update format
* sort import
* flake8
* Add results
* full results
* align the table
* refactor to inherit
* default per gpu eval = 1
* Formatting
* Formatting
* isort
* modify readme
* Add check
* Fix format
* Fix format
* Doc strings
* ALBERT & BERT for sequence classification don't inherit from the original anymore
* Remove incorrect comments
* Remove incorrect comments
* Remove incorrect comments
* Sync up with new code
* Sync up with new code
* Add a test
* Add a test
* Add a test
* Add a test
* Add a test
* Add a test
* Finishing up!
* Update hans data to be able to use Trainer
* Fixes
* Deal with tokenizer that don't have token_ids
* Clean up things
* Simplify data use
* Fix the input dict
* Formatting + proper path in README
* ner: add preprocessing script for examples that splits longer sentences
* ner: example shell scripts use local preprocessing now
* ner: add new example section for WNUT’17 NER task. Remove old English CoNLL-03 results
* ner: satisfy black and isort
* Glue task cleaup
* Enable writing cache to cache_dir in case dataset lives in readOnly
filesystem.
* Differentiate match vs mismatch for MNLI metrics.
* Style
* Fix pytype
* Fix type
* Use cache_dir in mnli mismatch eval dataset
* Small Tweaks
Co-authored-by: Julien Chaumond <chaumond@gmail.com>
* Kill model archive maps
* Fixup
* Also kill model_archive_map for MaskedBertPreTrainedModel
* Unhook config_archive_map
* Tokenizers: align with model id changes
* make style && make quality
* Fix CI
The option `--do_lower_case` is currently required by the uncased models (i.e., bert-base-uncased, bert-large-uncased).
Results:
BERT-BASE without --do_lower_case: 'exact': 73.83, 'f1': 82.22
BERT-BASE with --do_lower_case: 'exact': 81.02, 'f1': 88.34
* Adds predict stage for glue tasks, and generate result files which could be submitted to gluebenchmark.com website.
* Use Split enum + always output the label name
Co-authored-by: Julien Chaumond <chaumond@gmail.com>
* Distributed eval: SequentialDistributedSampler + gather all results
* For consistency only write to disk from world_master
Close https://github.com/huggingface/transformers/issues/4272
* Working distributed eval
* Hook into scripts
* Fix#3721 again
* TPU.mesh_reduce: stay in tensor space
Thanks @jysohn23
* Just a small comment
* whitespace
* torch.hub: pip install packaging
* Add test scenarii
* Add QA trainer example for TF
* Make data_dir optional
* Fix parameter logic
* Fix feature convert
* Update the READMEs to add the question-answering task
* Apply style
* Change 'sequence-classification' to 'text-classification' and prefix with 'eval' all the metric names
* Apply style
* Apply style
* Improvements to the wandb integration
* small reorg + no global necessary
* feat(trainer): log epoch and final metrics
* Simplify logging a bit
* Fixup
* Fix crash when just running eval
Co-authored-by: Chris Van Pelt <vanpelt@gmail.com>
Co-authored-by: Boris Dayma <boris.dayma@gmail.com>
* catch gpu len 1 set to gpu0
* Add mpc to trainer
* Add MPC for TF
* fix TF automodel for MPC and add Albert
* Apply style
* Fix import
* Note to self: double check
* Make shape None, None for datasetgenerator output shapes
* Add from_pt bool which doesnt seem to work
* Original checkpoint dir
* Fix docstrings for automodel
* Update readme and apply style
* Colab should probably not be from users
* Colabs should probably not be from users
* Add colab
* Update README.md
* Update README.md
* Cleanup __intit__
* Cleanup flake8 trailing comma
* Update src/transformers/training_args_tf.py
* Update src/transformers/modeling_tf_auto.py
Co-authored-by: Viktor Alm <viktoralm@pop-os.localdomain>
Co-authored-by: Julien Chaumond <chaumond@gmail.com>
* Created using Colaboratory
* [examples] reorganize files
* remove run_tpu_glue.py as superseded by TPU support in Trainer
* Bugfix: int, not tuple
* move files around
* First commit to add a TF version of the trainer.
* Make the TF trainer closer to what looks the PT trainer
* Refactoring common code between the PT and TF trainer into an util file.
* Some bugfix + better similarity with the PT trainer
* Add missing class in transformers init
* Bugfix over prediction + use classification report instead of simple metrics
* Fix name error
* Fix optimization tests + style
* Apply style
* Several bugfix for multi-gpu training
* Apply style
* Apply style
* Add glue example for the TF trainer
* Several bugix + address the reviews
* Fix on the TF training args file
* Add a debug mode
* Bugfix in utils_ner.py when segment_ids is None
* Apply style
* Apply style
* Add TPU strategy
* Fix selection strategy
* doc
* [tests] Add sample files for a regression task
* [HUGE] Trainer
* Feedback from @sshleifer
* Feedback from @thomwolf + logging tweak
* [file_utils] when downloading concurrently, get_from_cache will use the cached file for subsequent processes
* [glue] Use default max_seq_length of 128 like before
* [glue] move DataTrainingArguments around
* [ner] Change interface of InputExample, and align run_{tf,pl}
* Re-align the pl scripts a little bit
* ner
* [ner] Add integration test
* Fix language_modeling with API tweak
* [ci] Tweak loss target
* Don't break console output
* amp.initialize: model must be on right device before
* [multiple-choice] update for Trainer
* Re-align to 827d6d6ef0
* First pass on utility classes and python tokenizers
* finishing cleanup pass
* style and quality
* Fix tests
* Updating following @mfuntowicz comment
* style and quality
* Fix Roberta
* fix batch_size/seq_length inBatchEncoding
* add alignement methods + tests
* Fix OpenAI and Transfo-XL tokenizers
* adding trim_offsets=True default for GPT2 et RoBERTa
* style and quality
* fix tests
* add_prefix_space in roberta
* bump up tokenizers to rc7
* style
* unfortunately tensorfow does like these - removing shape/seq_len for now
* Update src/transformers/tokenization_utils.py
Co-Authored-By: Stefan Schweter <stefan@schweter.it>
* Adding doc and docstrings
* making flake8 happy
Co-authored-by: Stefan Schweter <stefan@schweter.it>
* Refactored use of newstest2013 to newstest2014. Fixed bug where argparse consumed first command line argument as model_size argument rather than using default model_size by forcing explicit --model_size flag inclusion
* More pythonic file handling through 'with' context
* COSMETIC - ran Black and isort
* Fixed reference to number of lines in newstest2014
* Fixed failing test. More pythonic file handling
* finish PR from tholiao
* remove outcommented lines
* make style
* make isort happy
Co-authored-by: Thomas Liao <tholiao@gmail.com>
* remove output_past from pt
* make style
* add optional input length for gpt2
* add use cache to prepare input
* save memory in gpt2
* correct gpt2 test inputs
* make past input optional for gpt2
* finish use_cache for all models
* make style
* delete modeling_gpt2 change in test file
* correct docstring
* correct is true statements for gpt2
* Initial commit to get BERT + run_glue.py on TPU
* Add README section for TPU and address comments.
* Cleanup TPU bits from run_glue.py (#3)
TPU runner is currently implemented in:
https://github.com/pytorch-tpu/transformers/blob/tpu/examples/run_glue_tpu.py.
We plan to upstream this directly into `huggingface/transformers`
(either `master` or `tpu`) branch once it's been more thoroughly tested.
* Cleanup TPU bits from run_glue.py
TPU runner is currently implemented in:
https://github.com/pytorch-tpu/transformers/blob/tpu/examples/run_glue_tpu.py.
We plan to upstream this directly into `huggingface/transformers`
(either `master` or `tpu`) branch once it's been more thoroughly tested.
* No need to call `xm.mark_step()` explicitly (#4)
Since for gradient accumulation we're accumulating on batches from
`ParallelLoader` instance which on next() marks the step itself.
* Resolve R/W conflicts from multiprocessing (#5)
* Add XLNet in list of models for `run_glue_tpu.py` (#6)
* Add RoBERTa to list of models in TPU GLUE (#7)
* Add RoBERTa and DistilBert to list of models in TPU GLUE (#8)
* Use barriers to reduce duplicate work/resources (#9)
* Shard eval dataset and aggregate eval metrics (#10)
* Shard eval dataset and aggregate eval metrics
Also, instead of calling `eval_loss.item()` every time do summation with
tensors on device.
* Change defaultdict to float
* Reduce the pred, label tensors instead of metrics
As brought up during review some metrics like f1 cannot be aggregated
via averaging. GLUE task metrics depends largely on the dataset, so
instead we sync the prediction and label tensors so that the metrics can
be computed accurately on those instead.
* Only use tb_writer from master (#11)
* Apply huggingface black code formatting
* Style
* Remove `--do_lower_case` as example uses cased
* Add option to specify tensorboard logdir
This is needed for our testing framework which checks regressions
against key metrics writtern by the summary writer.
* Using configuration for `xla_device`
* Prefix TPU specific comments.
* num_cores clarification and namespace eval metrics
* Cache features file under `args.cache_dir`
Instead of under `args.data_dir`. This is needed as our test infra uses
data_dir with a read-only filesystem.
* Rename `run_glue_tpu` to `run_tpu_glue`
Co-authored-by: LysandreJik <lysandre.debut@reseau.eseo.fr>
* [examples] Generate argparsers from type hints on dataclasses
* [HfArgumentParser] way simpler API
* Restore run_language_modeling.py for easier diff
* [HfArgumentParser] final tweaks from code review
* Big cleanup of `glue_convert_examples_to_features`
* Use batch_encode_plus
* Cleaner wrapping of glue_convert_examples_to_features for TF
@lysandrejik
* Cleanup syntax, thanks to @mfuntowicz
* Raise explicit error in case of user error
* Fix RoBERTa/XLNet Pad Token in run_multiple_choice.py
`convert_examples_to_fes atures` sets `pad_token=0` by default, which is correct for BERT but incorrect for RoBERTa (`pad_token=1`) and XLNet (`pad_token=5`). I think the other arguments to `convert_examples_to_features` are correct, but it might be helpful if someone checked who is more familiar with this part of the codebase.
* Simplifying change to match recent commits
* Using loaded checkpoint with --do_predict
Without this fix, I'm getting near-random validation performance for a trained model, and the validation performance differs per validation run. I think this happens since the `model` variable isn't set with the loaded checkpoint, so I'm using a randomly initialized model. Looking at the model activations, they differ each time I run evaluation (but they don't with this fix).
* Update checkpoint loading
* Fixing model loading
* Update the NER TF script to remove the softmax and make the pad token label id to -1
* Reformat the quality and style
Co-authored-by: Julien Plu <julien.plu@adevinta.com>
* force bleu
* fix wrong file name
* rename file
* different filenames for each example test
* test files should clean up after themselves
* test files should clean up after themselves
* do not force bleu
* correct typo
* fix isort
* Use tokenizer.num_added_tokens to count number of added special_tokens instead of hardcoded numbers.
Signed-off-by: Morgan Funtowicz <morgan@huggingface.co>
* run_ner.py - Do not add a label to the labels_ids if word_tokens is empty.
This can happen when using bert-base-multilingual-cased with an input containing an unique space.
In this case, the tokenizer will output just an empty word_tokens thus leading to an non-consistent behavior
over the labels_ids tokens adding one more tokens than tokens vector.
Signed-off-by: Morgan Funtowicz <morgan@huggingface.co>
* ✨ Alter base pl transformer to use automodels
* 🐛 Add batch size env variable to function call
* 💄 Apply black code style from Makefile
* 🚚 Move lightning base out of ner directory
* ✨ Add lightning glue example
* 💄 self
* move _feature_file to base class
* ✨ Move eval logging to custom callback
* 💄 Apply black code style
* 🐛 Add parent to pythonpath, remove copy command
* 🐛 Add missing max_length kwarg
* memory benchmark rss
* have both forward pass and line-by-line mem tracing
* cleaned up tracing
* refactored and cleaning up API
* no f-strings yet...
* add GPU mem logging
* fix GPU memory monitoring
* style and quality
* clean up and doc
* update with comments
* Switching to python 3.6+
* fix quality
* Rename and improve example
* Add test
* slightly faster test
* style
* This breaks remy prolly
* shorter test string
* no slow
* newdir structure
* New tree
* Style
* shorter
* docs
* clean
* Attempt future import
* more import hax
* * Added support for Albert when fine-tuning for NER
* Added support for Albert in NER pipeline
* Added command-line options to examples/ner/run_ner.py to better control tokenization
* Added class AlbertForTokenClassification
* Changed output for NerPipeline to use .convert_ids_to_tokens(...) instead of .decode(...) to better reflect tokens
* Added ,
* Now passes style guide enforcement
* Changes from reviews.
* Code now passes style enforcement
* Added test for AlbertForTokenClassification
* Added test for AlbertForTokenClassification
* add preprocessing to add space before punctuation for transfo_xl
* improve warning messages
* make style
* compile regex at instantination of tokenizer object
* Added support for Albert in NER pipeline
* Added command-line options to examples/ner/run_ner.py to better control tokenization
* Added class AlbertForTokenClassification
* Changed output for NerPipeline to use .convert_ids_to_tokens(...) instead of .decode(...) to better reflect tokens
* 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>
* pass langs parameter to certain XLM models
Adding an argument that specifies the language the SQuAD dataset is in so language-sensitive XLMs (e.g. `xlm-mlm-tlm-xnli15-1024`) don't default to language `0`.
Allows resolution of issue #1799 .
* fixing from `make style`
* fixing style (again)
Use -e only in docs targeted at contributors.
If a user copy-pastes command line with [--editable], they will hit
an error. If they don't know the --editable option, we're giving them
a choice to make before they can move forwards, but this isn't a choice
they need to make right now.
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.
These libraries aren't always installed in the virtual environment where
isort is running. Declaring them properly avoids mixing these
third-party imports with local imports.
This change is mostly autogenerated with:
$ python -m autoflake --in-place --recursive --remove-all-unused-imports --ignore-init-module-imports examples templates transformers utils hubconf.py setup.py
I made minor changes in the generated diff.
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:
$ 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.
* Small clarification
Matches line 431 to line 435 for additional clarity and consistency.
* Fixed minor typo
The letter "s" was previously omitted from the word "docstrings".
We currently save the pretrained_weights of the encoder and decoder in
two separate directories `encoder` and `decoder`. However, for the
`from_pretrained` function to operate with automodels we need to
specify the type of model in the path to the weights.
The path to the encoder/decoder weights is handled by the
`PreTrainedEncoderDecoder` class in the `save_pretrained` function. Sice
there is no easy way to infer the type of model that was initialized for
the encoder and decoder we add a parameter `model_type` to the function.
This is not an ideal solution as it is error prone, and the model type
should be carried by the Model classes somehow.
This is a temporary fix that should be changed before merging.
When evaluating, shouldn't we always use the SequentialSampler instead of DistributedSampler? Evaluation only runs on 1 GPU no matter what, so if you use the DistributedSampler with N GPUs, I think you'll only evaluate on 1/N of the evaluation set. That's at least what I'm finding when I run an older/modified version of this repo.
the data provided by Li Dong et al. were already tokenized, which means
that they are not compatible with all the models in the library. We
thus process the raw data directly and tokenize them using the models'
tokenizers.
We write a function to load an preprocess the CNN/Daily Mail dataset as
provided by Li Dong et al. The issue is that this dataset has already
been tokenized by the authors, so we actually need to find the original,
plain-text dataset if we want to apply it to all models.
adding conversion script
adding first draft of modeling & tokenization
adding placeholder for test files
bunch of changes
registering the tokenizer/model/etc
tests
change link; something is very VERY wrong here
weird end-of-word thingy going on
i think the tokenization works now ; wrote the unit tests
overall structure works;load w next
the monster is alive!
works after some cleanup as well
adding emacs autosave to gitignore
currently only supporting the 48 layer one; seems to infer fine on my macbook
cleanup
fixing some documentation
fixing some documentation
tests passing?
now works on CUDA also
adding greedy?
adding greedy sampling
works well
# Please enter a commit message to explain why this merge is necessary,
# especially if it merges an updated upstream into a topic branch.
#
# Lines starting with '#' will be ignored, and an empty message aborts
# the commit.
