Branch:
master
master
11 Commits (master)
11 Commits (master)
| Author | SHA1 | Message | Date |
|---|---|---|---|
 yuze.zyz
|
bb5512d1ab |
[to #42322933] Refactor NLP and fix some user feedbacks
1. Abstract keys of dicts needed by nlp metric classes into the init method
2. Add Preprocessor.save_pretrained to save preprocessor information
3. Abstract the config saving function, which can lead to normally saving in the direct call of from_pretrained, and the modification of cfg one by one when training.
4. Remove SbertTokenizer and VecoTokenizer, use transformers' tokenizers instead
5. Use model/preprocessor's from_pretrained in all nlp pipeline classes.
6. Add model_kwargs and preprocessor_kwargs in all nlp pipeline classes
7. Add base classes for fill-mask and text-classification preprocessor, as a demo for later changes
8. Fix user feedback: Re-train the model in continue training scenario
9. Fix user feedback: Too many checkpoint saved
10. Simplify the nlp-trainer
11. Fix user feedback: Split the default trainer's __init__ method, which makes user easier to override
12. Add safe_get to Config class
---------------------------- Another refactor from version 36 -------------------------
13. Name all nlp transformers' preprocessors from TaskNamePreprocessor to TaskNameTransformersPreprocessor, for example:
TextClassificationPreprocessor -> TextClassificationTransformersPreprocessor
14. Add a base class per task for all nlp tasks' preprocessors which has at least two sub-preprocessors
15. Add output classes of nlp models
16. Refactor the logic for token-classification
17. Fix bug: checkpoint_hook does not support pytorch_model.pt
18. Fix bug: Pipeline name does not match with task name, so inference will not succeed after training
NOTE: This is just a stop bleeding solution, the root cause is the uncertainty of the relationship between models and pipelines
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/10723513
* add save_pretrained to preprocessor
* save preprocessor config in hook
* refactor label-id mapping fetching logic
* test ok on sentence-similarity
* run on finetuning
* fix bug
* pre-commit passed
* fix bug
* Merge branch 'master' into feat/refactor_config
# Conflicts:
# modelscope/preprocessors/nlp/nlp_base.py
* add params to init
* 1. support max ckpt num 2. support ignoring others but bin file in continue training 3. add arguments to some nlp metrics
* Split trainer init impls to overridable methods
* remove some obsolete tokenizers
* unfinished
* support input params in pipeline
* fix bugs
* fix ut bug
* fix bug
* fix ut bug
* fix ut bug
* fix ut bug
* add base class for some preprocessors
* Merge commit '379867739548f394d0fa349ba07afe04adf4c8b6' into feat/refactor_config
* compatible with old code
* fix ut bug
* fix ut bugs
* fix bug
* add some comments
* fix ut bug
* add a requirement
* fix pre-commit
* Merge commit '0451b3d3cb2bebfef92ec2c227b2a3dd8d01dc6a' into feat/refactor_config
* fixbug
* Support function type in registry
* fix ut bug
* fix bug
* Merge commit '5f719e542b963f0d35457e5359df879a5eb80b82' into feat/refactor_config
# Conflicts:
# modelscope/pipelines/nlp/multilingual_word_segmentation_pipeline.py
# modelscope/pipelines/nlp/named_entity_recognition_pipeline.py
# modelscope/pipelines/nlp/word_segmentation_pipeline.py
# modelscope/utils/hub.py
* remove obsolete file
* rename init args
* rename params
* fix merge bug
* add default preprocessor config for ner-model
* move a method a util file
* remove unused config
* Fix a bug in pbar
* bestckptsaver:change default ckpt numbers to 1
* 1. Add assert to max_epoch 2. split init_dist and get_device 3. change cmp func name
* Fix bug
* fix bug
* fix bug
* unfinished refactoring
* unfinished
* uw
* uw
* uw
* uw
* Merge branch 'feat/refactor_config' into feat/refactor_trainer
# Conflicts:
# modelscope/preprocessors/nlp/document_segmentation_preprocessor.py
# modelscope/preprocessors/nlp/faq_question_answering_preprocessor.py
# modelscope/preprocessors/nlp/relation_extraction_preprocessor.py
# modelscope/preprocessors/nlp/text_generation_preprocessor.py
* uw
* uw
* unify nlp task outputs
* uw
* uw
* uw
* uw
* change the order of text cls pipeline
* refactor t5
* refactor tg task preprocessor
* fix
* unfinished
* temp
* refactor code
* unfinished
* unfinished
* unfinished
* unfinished
* uw
* Merge branch 'feat/refactor_config' into feat/refactor_trainer
* smoke test pass
* ut testing
* pre-commit passed
* Merge branch 'master' into feat/refactor_config
# Conflicts:
# modelscope/models/nlp/bert/document_segmentation.py
# modelscope/pipelines/nlp/__init__.py
# modelscope/pipelines/nlp/document_segmentation_pipeline.py
* merge master
* unifnished
* Merge branch 'feat/fix_bug_pipeline_name' into feat/refactor_config
* fix bug
* fix ut bug
* support ner batch inference
* fix ut bug
* fix bug
* support batch inference on three nlp tasks
* unfinished
* fix bug
* fix bug
* Merge branch 'master' into feat/refactor_config
# Conflicts:
# modelscope/models/base/base_model.py
# modelscope/pipelines/nlp/conversational_text_to_sql_pipeline.py
# modelscope/pipelines/nlp/dialog_intent_prediction_pipeline.py
# modelscope/pipelines/nlp/dialog_modeling_pipeline.py
# modelscope/pipelines/nlp/dialog_state_tracking_pipeline.py
# modelscope/pipelines/nlp/document_segmentation_pipeline.py
# modelscope/pipelines/nlp/faq_question_answering_pipeline.py
# modelscope/pipelines/nlp/feature_extraction_pipeline.py
# modelscope/pipelines/nlp/fill_mask_pipeline.py
# modelscope/pipelines/nlp/information_extraction_pipeline.py
# modelscope/pipelines/nlp/named_entity_recognition_pipeline.py
# modelscope/pipelines/nlp/sentence_embedding_pipeline.py
# modelscope/pipelines/nlp/summarization_pipeline.py
# modelscope/pipelines/nlp/table_question_answering_pipeline.py
# modelscope/pipelines/nlp/text2text_generation_pipeline.py
# modelscope/pipelines/nlp/text_classification_pipeline.py
# modelscope/pipelines/nlp/text_error_correction_pipeline.py
# modelscope/pipelines/nlp/text_generation_pipeline.py
# modelscope/pipelines/nlp/text_ranking_pipeline.py
# modelscope/pipelines/nlp/token_classification_pipeline.py
# modelscope/pipelines/nlp/word_segmentation_pipeline.py
# modelscope/pipelines/nlp/zero_shot_classification_pipeline.py
# modelscope/trainers/nlp_trainer.py
* pre-commit passed
* fix bug
* Merge branch 'master' into feat/refactor_config
# Conflicts:
# modelscope/preprocessors/__init__.py
* fix bug
* fix bug
* fix bug
* fix bug
* fix bug
* fixbug
* pre-commit passed
* fix bug
* fixbug
* fix bug
* fix bug
* fix bug
* fix bug
* self review done
* fixbug
* fix bug
* fix bug
* fix bugs
* remove sub-token offset mapping
* fix name bug
* add some tests
* 1. support batch inference of text-generation,text2text-generation,token-classification,text-classification 2. add corresponding UTs
* add old logic back
* tmp save
* add tokenize by words logic back
* move outputs file back
* revert veco token-classification back
* fix typo
* Fix description
* Merge commit '4dd99b8f6e4e7aefe047c68a1bedd95d3ec596d6' into feat/refactor_config
* Merge branch 'master' into feat/refactor_config
# Conflicts:
# modelscope/pipelines/builder.py
|
2 years ago |
|
yuze.zyz
|
4b7e8e89aa |
[to #42322933] Fix some bugs when downgrade the version of some dependencies
1. Fix bug in model exporting
2. Skip some long trainings in test level 2
3. Refine some comments
4. Fix a bug that mode is not correct when saving checkpoints
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/10564716
|
2 years ago |
|
yuze.zyz
|
212cf53318 |
[to #42322933] Fix some bugs
1. Add F1 score to sequence classification metric
2. Fix a bug that the evaluate method in trainer does not support a pure pytorch_model.bin
3. Fix a bug in evaluation of veco trainer
4. Add some tips if lr_scheduler in the trainer needs a higher version torch
5. Add some comments
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/10532230
|
2 years ago |
|
yuze.zyz
|
605cd7f44a |
[to #42322933] NLP 1030 Refactor
Features:
1. Refactor the directory structure of nlp models. All model files are placed into either the model folder or the task_model folder
2. Refactor all the comments to google style
3. Add detail comments to important tasks and nlp models, to list the description of the model, and its preprocessor&trainer
4. Model Exporting now supports a direct all to TorchModelExporter(no need to derive from it)
5. Refactor model save_pretrained method to support direct running(independent from trainer)
6. Remove the judgement of Model in the pipeline base class, to support outer register models running in our pipelines
7. Nlp trainer now has a NLPTrainingArguments class , user can pass arguments into the dataclass, and use it as a normal cfg_modify_fn, to simplify the operation of modify cfg.
8. Merge the BACKBONES and the MODELS, so user can get a backbone with the Model.from_pretrained call
9. Model.from_pretrained now support a task argument, so user can use a backbone and load it with a specific task class.
10. Support Preprocessor.from_pretrained method
11. Add standard return classes to important nlp tasks, so some of the pipelines and the models are independent now, the return values of the models will always be tensors, and the pipelines will take care of the conversion to numpy and the following stuffs.
12. Split the file of the nlp preprocessors, to make the dir structure more clear.
Bugs Fixing:
1. Fix a bug that lr_scheduler can be called earlier than the optimizer's step
2. Fix a bug that the direct call of Pipelines (not from pipeline(xxx)) throws error
3. Fix a bug that the trainer will not call the correct TaskDataset class
4. Fix a bug that the internal loading of dataset will throws error in the trainer class
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/10490585
|
2 years ago |
|
yuze.zyz
|
707cbef013 |
[to #42322933]Fix bug in daily UT
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/10491891 |
3 years ago |
|
yuze.zyz
|
acba1786b0 |
[to #42322933] Fix bug in UT daily
1. Fix bugs in daily test
2. Fix a bug that the updating of lr is before the first time of updating of optimizer
TODO this will still cause warnings when GA is above 1
3. Remove the judgement of mode in text-classification's preprocessor to fit the base trainer(Bug)
Update some regression bins to fit the preprocessor
4. Update the regression tool to let outer code modify atol and rtol
5. Add the default metric for text-classification task
6. Remove the useless ckpt conversion method in bert to avoid the requirement of tf when loading modeling_bert
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/10430764
|
3 years ago |
|
yuze.zyz
|
01e768503c |
[to #42322933] Fix random seed for trainer
1. Fix random seed for trainer and init it at the first line of init
2. Add a regress test for fixed training
3. Change the dataset 'dureader_robust_qg' to 'DuReader_robust-QG'
4. Change some datasets from loading hf.datasets to loading msdataset.load
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/10029509
|
3 years ago |
|
yuze.zyz
|
88d0804dcd |
[to #42322933] Add S4: child-tuning
1. add child-tuning optimizer and ut
2. fix a training bug which can cause interruption after cross-evaluation
3. move model.params from cfg to default args in build_optimizer to prevent the saving of params in save_pretrained
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9891963
|
3 years ago |
 zhangzhicheng.zzc
|
b94bb74f66 | [to #42322933]Add model.save_pretrained method and allow finetune results used by pipeline | 3 years ago |
|
zhangzhicheng.zzc
|
9d0b38b4e4 | [to #42322933] lazy load on trainer | 3 years ago |
|
yuze.zyz
|
21fa71baf0 |
[to #42322933] add/refactor nlp models source code and finetune
1. add sbert,veco,palm,space source code
2. support sbert sequence classification, token classification finetune
3. support veco sequence classification finetune
4. support palm nlg finetune
evaluation result: https://sheet.alibaba-inc.com/#/sheet/f7fdcc7f22bd5105 sheet:Maas
5. add ut for finetunes
6. add veco's taskdataset processor
7. add a common trainer for nlp, and a specific trainer for veco
8. merge some duplicate codes of models, preprocessors, pipelines
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9574105
* add basic class of hook&metrics
* pre-commit passed
* change some comments
* pre commit passed
* 1. remove accuracy's groups 2. remove useless hooks 3. simplify priorities
* pre-commit passed
* fix a comment
* Merge branch 'master' into finetune_hooks_metrics
# Conflicts:
# modelscope/metainfo.py
* pre-commit passed
* add basic class of hook&metrics
* pre-commit passed
* change some comments
* pre commit passed
* 1. remove accuracy's groups 2. remove useless hooks 3. simplify priorities
* pre-commit passed
* fix a comment
* Merge branch 'feat/finetune' of gitlab.alibaba-inc.com:Ali-MaaS/MaaS-lib into feat/finetune
* mv hooks related to modelscope/trainers/hooks
* mv priority back
* add torch mdoel base and test
* update hooks, trainer, import_util
* add torch epoch based trainer and dis utils
* add hooks
* fix warmup
* format code stype and fix warmup and add warmup unittest
* fix impls
* pre-commit check passed
* update hook and add EpochBasedTrainer
* add trainer unittest
* Merge branch 'feat/add_hooks' into feat/add_task
# Conflicts:
# modelscope/models/base_torch.py
# modelscope/trainers/hooks/hook.py
# modelscope/trainers/trainer.py
* update unittest name
* rewrite taskdataset to trainer
* fix trainer and add unittest
* add unittest
* code: run to forward
* run through... but ugly code
* arrange some cls
* fix some errs
* revert some mistakes
* init check in
* Merge branch 'feat/add_hooks' into feat/add_task
# Conflicts:
# modelscope/trainers/trainer.py
* test with bigger epoch and size
* add the default metrics class
* move build metrics code to a method
* merge add_task
* merge origin add_task
* add device initialization
* remove preprocessor arg for bool
* add task models
* move metric collect logic to metrics class
* pre-commit passed
* fix cr comments
* precommit passed
* add task models
* Merge remote-tracking branch 'origin/feat/add_task' into feat/backbone_head
* add comment
* change comment formats.
* fix comments
* fix ut bug
* fix comments
* add wrapper check
* fix comments
* pre commit passed
* fix cr comments
* solve a loop import problem
* fix ut bug
* fix ut errors
* change dummydataset to msdataset
* precommit passed
* merge add task
* backbone-head is build, model is not correctly loaded
* model load states matched
* result matched
* lint
* add veco/palm_v2 code
* merge master
* merge master success running
* add repr model name level
* Merge branch 'feat/veco_palm' into feat/finetune_sbert_veco
* model test for training
* add token-classification metric add formal ut
* fix running bug
* finetune and pipeline are working with backbone-head
* add nli
* add missing code
* finetune and pipeline are working with backbone-head
* Merge branch 'feat/backbone_head' of http://gitlab.alibaba-inc.com/Ali-MaaS/MaaS-lib into feat/backbone_head
* add a test repo for pr
* remove merge conflicted file
* remove merge conflicted file 1
* lint check
* import error
* none type bug fix
* forward input unpacking or dict bug
* move head into models, add build_backbone with registry, no base method
* merge master
* feat: 1. add interleave dataset method 2. support multiple dataset in trainer.build_dataset 3. support 3 sub tasks in sequence_classification task
* unfinished
* update the task model structure in NLP field
* merge master
* update by comments
* keep the default model id as current on production
* unfinished
* unfinished
* veco can run
* Merge remote-tracking branch 'origin/master' into feat/backbone_head
* add taskmodel for module management
* remove forward_input_is_dict
* unfinished
* token classification started
* update base model structure
* move space to backbone
* remove 'type' in build_from_cfg method
* test update
* bug fix
* on tesing, mess code
* Merge branch 'feat/backbone_head' into feat/refactor_nlp_730
# Conflicts:
# modelscope/metrics/builder.py
# modelscope/models/__init__.py
# modelscope/models/nlp/__init__.py
# modelscope/preprocessors/nlp.py
# modelscope/trainers/trainer.py
# requirements/multi-modal.txt
* add missing merge
* add sofa source code
* refactor
* add veco task dataset
* add veco task dataset
* pre-commit passed
* fix bug of log
* add some features
* merge master
* bug fix
* refine nlp models
* fix the training error
* unfinished
* refactor pipeline
* Merge branch 'feat/backbone_head' into feat/refactor_nlp_730
# Conflicts:
# modelscope/metrics/builder.py
# modelscope/models/nlp/__init__.py
# modelscope/models/nlp/backbones/structbert/modeling_sbert.py
# modelscope/models/nlp/palm_v2/palm_for_text_generation.py
# modelscope/preprocessors/base.py
# modelscope/preprocessors/nlp.py
# modelscope/trainers/trainer.py
* Merge commit 'ab04ceafc5453ce7daa9aa09e37a55f703072a10' into feat/refactor_nlp_730
# Conflicts:
# modelscope/metainfo.py
# modelscope/metrics/builder.py
# modelscope/models/__init__.py
# modelscope/models/base/base_torch_model.py
# modelscope/models/nlp/__init__.py
# modelscope/models/nlp/backbones/space/model/intent_unified_transformer.py
# modelscope/models/nlp/backbones/space/model/model_base.py
# modelscope/models/nlp/palm_v2/palm_for_text_generation.py
# modelscope/models/nlp/sbert_for_sequence_classification.py
# modelscope/models/nlp/sequence_classification.py
# modelscope/models/nlp/space/__init__.py
# modelscope/models/nlp/space_for_dialog_intent_prediction.py
# modelscope/models/nlp/space_for_dialog_modeling.py
# modelscope/models/nlp/space_for_dialog_state_tracking.py
# modelscope/models/nlp/task_model.py
# modelscope/pipelines/nlp/sentiment_classification_pipeline.py
# modelscope/preprocessors/base.py
# modelscope/preprocessors/nlp.py
# modelscope/trainers/trainer.py
* revert changes
* unify sentnece classification postprocess
* revert some changes, move some model files
* pipeline first case run through
* ws pipeline passed
* Merge branch 'feat/refactor_nlp_730' into feat/finetune_sbert_veco
* finetune
* revert code
* revert some code
* ws finetune started, only the accuracy is weird
* Merge branch 'feat/veco_taskdataset' into feat/finetune_sbert_veco
# Conflicts:
# modelscope/task_datasets/veco_dataset.py
# tests/taskdataset/test_veco_dataset.py
* veco+nli finetune started
* Merge branch 'master' into feat/finetune_sbert_veco
# Conflicts:
# modelscope/models/nlp/sbert_for_sequence_classification.py
# modelscope/models/nlp/sbert_for_token_classification.py
# modelscope/models/nlp/sbert_for_zero_shot_classification.py
# modelscope/models/nlp/space/space_for_dialog_intent_prediction.py
# modelscope/models/nlp/space/space_for_dialog_modeling.py
# modelscope/trainers/trainer.py
* add trainer for nlp
* trainer: dataset params passed into preprocessor
* test passed by nlptrainer
* fix some bugs
* fix some bugs
* add backbone/head subclass
* fix regression bugs
* fix bug in token-cls finetune
* support cfg modification
* fix bug
* fix bug
* update requirements
* add some comments and fix some t
* add some comments and revert a argument
* split to two test files
* revert code
* fixbug in precessor
(cherry picked from commit
|
3 years ago |