wassname/Castor
1
0
Fork 0
mirror of https://github.com/wassname/Castor.git synced 2026-08-20 12:00:37 +08:00
Code Issues Packages Projects Releases Wiki Activity
143 Commits 6 Branches 0 Tags
master
Commit Graph
143 Commits
This Branch
This Branch
All Branches
Author SHA1 Message Date
wassname 90477f2274 update to pytorch1.0.0 and torchtext 0.3.1 2019-02-23 14:33:40 +08:00
Ralph Tang 3855254870 Tidy up some code (#174) 2019-02-06 13:36:48 -05:00
Ashutosh-Adhikari 84225d208c Add char_quantize for AAPD (#173) 2019-01-29 09:52:20 -05:00
Victor Yang aec0826356 Add ESIM model (#169) 
* runnable

* update mask

* minor update

* minor update

* Update README.md

* fix multi GPU issue

* add visualize argument

* fix more comments, retab

* remove util
 2019-01-28 22:36:58 -05:00
Ashutosh-Adhikari d326e575c6 Add TAR and AR (#172) 
* Add TAR and AR
 2019-01-25 16:39:35 -05:00
Ashutosh-Adhikari dc086e895f Add document classification models and datasets (#171) 
* Add ReutersTrainer, ReutersEvaluator options in Factory classes

* Add Reuters to Kim-CNN command line arguments

* Fix SST dataset path according to changes in Kim-CNN args

The dataset path in args.py was made to point at the dataset folder rather than dataset/SST folder. Hence SST folder was added to paths in the SST dataset class

* Add Reuters dataset class, and support in __main__

* Add Reuters dataset trainers and evaluators

* Remove debug print statement in reuters_evaluator

* Fix rounding bug in reuters_trainer and reuters_evaluator

* Add LSTM for baseline text classification measurements

* Add eval metrics for lstm_baseline

* Set batch_first param in lstm_baseline

* Remove onnx args from lstm_baseline

* Pack padded sequences in LSTM_baseline

* Add TensorBoardX support for Reuters trainer

* Add Arxiv Academic Paper Dataset (AAPD)

* Add Hidden Bottleneck Layer to BiLSTM

* Fix packing of padded tensors in Reuters

* Add cmdline args for Hidden Bottleneck Layer for BiLSTM

* Include pre-padding lengths in AAPD dataset

* Remove duplication of preprocessing code in AAPD

* Remove batch_size condition in ReutersTrainer

* Add ignore_lengths option to ReutersTrainer and ReutersEvaluator

* Add AAPDCharQuantized and ReutersCharQuantized

* Rename Reuters_hierarchical to ReutersHierarchical

* Add CharacterCNN for document classification

* Update README.md for CharacterCNN

* Fix table in README.md for CharacterCNN

* Add AAPDHierarchical for HAN

* Update HAN for changes in Reuters dataset endpoints

* Fix bug in CharCNN when running on CPU

* Add AAPD dataset support for KimCNN

* Fix dataset paths for SST-1

* Fix dimensions of FC1 in CharCNN

* Add model checkpointing for Reuters based on F1

* Refactor LSTM baseline __main__

* Add precision, recall and F1 to Reuters evaluator

* Checkpoint only at the end of an epoch for ReutersTrainer

Add detailed log printing for dev evaluations

* Fix log_template and dev_log_template in ReutersTrainer

* Add IMDB dataset

* Fix duplicate printing of header in ReutersTrainer

* Add support for single_label datasets in ReutersTrainer

* Add support for IMDB dataset in lstm_baseline and lstm_reg

* Fix evaluator call in main method of HAN

* Add IMDB for HAN

* Fix for single_label

* Fix evaluate_dataset method for single_label datasets

* Reduce default patience to 5 epochs before early stopping

* Revert change to save_state rather than the entire model

* Add Yelp 2018 dataset

* Integrate Yelp2018 with LSTM baseline

* Replace Yelp2018 with Yelp2014 dataset

* Add Yelp2014 to LSTM Baseline

* Integrate Yelp14 into LSTM Regularization

* Remove dropout in HBL for LSTM Baseline and Reg

* Add Yelp for HAN

* Fix the saving issue for HAN

* Fix loading for HAN

* Fix typo in ReutersEvaluator

* Print to STDOUT rather than logger

* Print XML-CNN eval to STDOUT rather than logger

* Update max_length for IMDB dataset

* Add single_label support for char_cnn

* Fix evaluation method for char_cnn

* Remove unwanted parameters from ReutersTrainer and ReutersEval

* Fix code formatting in lstm_reg/args

* Add support for IMDB and Yelp in KimCNN

* Fix single_label incorporation

* Remove unnecessary conditions

* Fix num_classes in Yelp2014

* Add single_label support for XML-CNN

* Fix call to evaluator in XML-CNN

* Address PEP8 issues

* Address PEP8 issues

* Address PEP8 issues

* Address PEP8 issues
 2019-01-25 13:02:37 -05:00
Victor Yang 57f53a81b5 Add SSE model (#168) 
* runnable

* add util file

* update readme

* update final layer and add model name

* update argument

* update readme, delete useless args

* fix comments

* fix more comments
 2018-12-17 21:04:37 -05:00
Victor Yang 7c3c15649d add DecAtt model (#170) 
* add DecAtt model

* update readme, add dropout

* fix more comments

* add trecqa, wikiqa results

* remove extraneous comment
 2018-12-17 21:03:29 -05:00
Victor Yang 56e29c886e add seperator (#167) 2018-12-03 18:41:32 -05:00
Ralph Tang f510fa6bae Fix VDPWI focus cube padding (#162) 2018-11-12 19:35:11 -05:00
Achyudh Ram 951df4ada5 Neural Document Classification (#159) 
* Add ReutersTrainer, ReutersEvaluator options in Factory classes

* Add Reuters to Kim-CNN command line arguments

* Fix SST dataset path according to changes in Kim-CNN args

The dataset path in args.py was made to point at the dataset folder rather than dataset/SST folder. Hence SST folder was added to paths in the SST dataset class

* Add Reuters dataset class, and support in __main__

* Add Reuters dataset trainers and evaluators

* Remove debug print statement in reuters_evaluator

* Fix rounding bug in reuters_trainer and reuters_evaluator

* Add LSTM for baseline text classification measurements

* Add eval metrics for lstm_baseline

* Set batch_first param in lstm_baseline

* Remove onnx args from lstm_baseline

* Pack padded sequences in LSTM_baseline

* Add TensorBoardX support for Reuters trainer

* Add Arxiv Academic Paper Dataset (AAPD)

* Add Hidden Bottleneck Layer to BiLSTM

* Fix packing of padded tensors in Reuters

* Add cmdline args for Hidden Bottleneck Layer for BiLSTM

* Include pre-padding lengths in AAPD dataset

* Remove duplication of preprocessing code in AAPD

* Remove batch_size condition in ReutersTrainer

* Add ignore_lengths option to ReutersTrainer and ReutersEvaluator

* Add AAPDCharQuantized and ReutersCharQuantized

* Rename Reuters_hierarchical to ReutersHierarchical

* Add CharacterCNN for document classification

* Update README.md for CharacterCNN

* Fix table in README.md for CharacterCNN

* Add AAPDHierarchical for HAN

* Update HAN for changes in Reuters dataset endpoints

* Fix bug in CharCNN when running on CPU

* Add AAPD dataset support for KimCNN

* Fix dataset paths for SST-1

* Fix dimensions of FC1 in CharCNN

* Add model checkpointing for Reuters based on F1

* Refactor LSTM baseline __main__

* Add precision, recall and F1 to Reuters evaluator

* Checkpoint only at the end of an epoch for ReutersTrainer

Add detailed log printing for dev evaluations

* Fix log_template and dev_log_template in ReutersTrainer

* Add IMDB dataset

* Add support for single_label datasets in ReutersTrainer

* Add support for IMDB dataset in lstm_baseline and lstm_reg
 2018-11-10 23:07:11 -05:00
Ashutosh-Adhikari 3dd3ced661 Fix HAN for batch_size 1 (#161) 2018-11-09 21:37:49 -05:00
Ashutosh-Adhikari 97bdaec5bb Add AAPD for XML_CNN (#160) 
* Add AAPD for XMLCNN

* Add kwargs for XML
 2018-11-09 20:53:43 -05:00
Achyudh Ram addc4506d1 Add model checkpointing to ReutersTrainer (#158) 
* Add ReutersTrainer, ReutersEvaluator options in Factory classes

* Add Reuters to Kim-CNN command line arguments

* Fix SST dataset path according to changes in Kim-CNN args

The dataset path in args.py was made to point at the dataset folder rather than dataset/SST folder. Hence SST folder was added to paths in the SST dataset class

* Add Reuters dataset class, and support in __main__

* Add Reuters dataset trainers and evaluators

* Remove debug print statement in reuters_evaluator

* Fix rounding bug in reuters_trainer and reuters_evaluator

* Add LSTM for baseline text classification measurements

* Add eval metrics for lstm_baseline

* Set batch_first param in lstm_baseline

* Remove onnx args from lstm_baseline

* Pack padded sequences in LSTM_baseline

* Add TensorBoardX support for Reuters trainer

* Add Arxiv Academic Paper Dataset (AAPD)

* Add Hidden Bottleneck Layer to BiLSTM

* Fix packing of padded tensors in Reuters

* Add cmdline args for Hidden Bottleneck Layer for BiLSTM

* Include pre-padding lengths in AAPD dataset

* Remove duplication of preprocessing code in AAPD

* Remove batch_size condition in ReutersTrainer

* Add ignore_lengths option to ReutersTrainer and ReutersEvaluator

* Add AAPDCharQuantized and ReutersCharQuantized

* Rename Reuters_hierarchical to ReutersHierarchical

* Add CharacterCNN for document classification

* Update README.md for CharacterCNN

* Fix table in README.md for CharacterCNN

* Add AAPDHierarchical for HAN

* Update HAN for changes in Reuters dataset endpoints

* Fix bug in CharCNN when running on CPU

* Add AAPD dataset support for KimCNN

* Fix dataset paths for SST-1

* Fix dimensions of FC1 in CharCNN

* Add model checkpointing for Reuters based on F1

* Refactor LSTM baseline __main__

* Add precision, recall and F1 to Reuters evaluator
 2018-11-07 17:15:30 -05:00
Ashutosh-Adhikari 91ed6261db Add regularization modules for LSTM baseline (#156) 
* Add Regularization Modules for LSTM

* Update Reuters Trainer and Evalueator for regularization

* Remove unnecessary comments

* Comply with PEP8

* Comply import order with PEP8

* Fix typos in README.md

* Comply with PEP8

* Add BSD 3-Clause Licence

* Remove deprecated call to Variable for PyTorch 0.4

* Update dataset selection in main

* Remove block comments
 2018-11-06 10:48:28 -05:00
Achyudh Ram f0a5c370bc Fix KimCNN for SST, AAPD datasets (#157) 
* Add ReutersTrainer, ReutersEvaluator options in Factory classes

* Add Reuters to Kim-CNN command line arguments

* Fix SST dataset path according to changes in Kim-CNN args

The dataset path in args.py was made to point at the dataset folder rather than dataset/SST folder. Hence SST folder was added to paths in the SST dataset class

* Add Reuters dataset class, and support in __main__

* Add Reuters dataset trainers and evaluators

* Remove debug print statement in reuters_evaluator

* Fix rounding bug in reuters_trainer and reuters_evaluator

* Add LSTM for baseline text classification measurements

* Add eval metrics for lstm_baseline

* Set batch_first param in lstm_baseline

* Remove onnx args from lstm_baseline

* Pack padded sequences in LSTM_baseline

* Add TensorBoardX support for Reuters trainer

* Add Arxiv Academic Paper Dataset (AAPD)

* Add Hidden Bottleneck Layer to BiLSTM

* Fix packing of padded tensors in Reuters

* Add cmdline args for Hidden Bottleneck Layer for BiLSTM

* Include pre-padding lengths in AAPD dataset

* Remove duplication of preprocessing code in AAPD

* Remove batch_size condition in ReutersTrainer

* Add ignore_lengths option to ReutersTrainer and ReutersEvaluator

* Add AAPDCharQuantized and ReutersCharQuantized

* Rename Reuters_hierarchical to ReutersHierarchical

* Add CharacterCNN for document classification

* Update README.md for CharacterCNN

* Fix table in README.md for CharacterCNN

* Add AAPDHierarchical for HAN

* Update HAN for changes in Reuters dataset endpoints

* Fix bug in CharCNN when running on CPU

* Add AAPD dataset support for KimCNN

* Fix dataset paths for SST-1
 2018-11-05 15:42:28 -05:00
Achyudh Ram cc275f6bde Add CharacterCNN for Document Classification (#155) 
* Add ReutersTrainer, ReutersEvaluator options in Factory classes

* Add Reuters to Kim-CNN command line arguments

* Fix SST dataset path according to changes in Kim-CNN args

The dataset path in args.py was made to point at the dataset folder rather than dataset/SST folder. Hence SST folder was added to paths in the SST dataset class

* Add Reuters dataset class, and support in __main__

* Add Reuters dataset trainers and evaluators

* Remove debug print statement in reuters_evaluator

* Fix rounding bug in reuters_trainer and reuters_evaluator

* Add LSTM for baseline text classification measurements

* Add eval metrics for lstm_baseline

* Set batch_first param in lstm_baseline

* Remove onnx args from lstm_baseline

* Pack padded sequences in LSTM_baseline

* Add TensorBoardX support for Reuters trainer

* Add Arxiv Academic Paper Dataset (AAPD)

* Add Hidden Bottleneck Layer to BiLSTM

* Fix packing of padded tensors in Reuters

* Add cmdline args for Hidden Bottleneck Layer for BiLSTM

* Include pre-padding lengths in AAPD dataset

* Remove duplication of preprocessing code in AAPD

* Remove batch_size condition in ReutersTrainer

* Add ignore_lengths option to ReutersTrainer and ReutersEvaluator

* Add AAPDCharQuantized and ReutersCharQuantized

* Rename Reuters_hierarchical to ReutersHierarchical

* Add CharacterCNN for document classification

* Update README.md for CharacterCNN

* Fix table in README.md for CharacterCNN

* Add AAPDHierarchical for HAN

* Update HAN for changes in Reuters dataset endpoints

* Fix bug in CharCNN when running on CPU
 2018-10-28 19:01:54 -04:00
Achyudh Ram 6daa5a128f Replication of STOA for Reuters Dataset (#152) 
* Add ReutersTrainer, ReutersEvaluator options in Factory classes

* Add Reuters to Kim-CNN command line arguments

* Fix SST dataset path according to changes in Kim-CNN args

The dataset path in args.py was made to point at the dataset folder rather than dataset/SST folder. Hence SST folder was added to paths in the SST dataset class

* Add Reuters dataset class, and support in __main__

* Add Reuters dataset trainers and evaluators

* Remove debug print statement in reuters_evaluator

* Fix rounding bug in reuters_trainer and reuters_evaluator

* Add LSTM for baseline text classification measurements

* Add eval metrics for lstm_baseline

* Set batch_first param in lstm_baseline

* Remove onnx args from lstm_baseline

* Pack padded sequences in LSTM_baseline

* Add TensorBoardX support for Reuters trainer

* Add Arxiv Academic Paper Dataset (AAPD)

* Add Hidden Bottleneck Layer to BiLSTM

* Fix packing of padded tensors in Reuters

* Add cmdline args for Hidden Bottleneck Layer for BiLSTM

* Include pre-padding lengths in AAPD dataset

* Remove duplication of preprocessing code in AAPD

* Remove batch_size condition in ReutersTrainer
 2018-10-26 19:10:19 -04:00
Ashutosh-Adhikari 650882fb6e Add HAN and XML_CNN for Doc Classification (#154) 
* Add Reuters option in common.dataset

* Add Reuters option in common.dataset

* Add HAN model

* Add XML-CNN

* Add HAN

* Add Hierarchical tokenization for Reuters

* Add README for HAN

* Add XML Readme

* Update HAN Readme
 2018-10-25 15:45:13 -04:00
Achyudh Ram ed4f01852e Baseline LSTM implementation (#150) 
* Add ReutersTrainer, ReutersEvaluator options in Factory classes

* Add Reuters to Kim-CNN command line arguments

* Fix SST dataset path according to changes in Kim-CNN args

The dataset path in args.py was made to point at the dataset folder rather than dataset/SST folder. Hence SST folder was added to paths in the SST dataset class

* Add Reuters dataset class, and support in __main__

* Add Reuters dataset trainers and evaluators

* Remove debug print statement in reuters_evaluator

* Fix rounding bug in reuters_trainer and reuters_evaluator

* Add LSTM for baseline text classification measurements

* Add eval metrics for lstm_baseline

* Set batch_first param in lstm_baseline

* Remove onnx args from lstm_baseline
 2018-10-11 03:04:23 -04:00
Linqing Liu 7dec34a106 Add sts2014 and quora evaluators in common/evaluators/ (#151) 
* add SNLI dataset

* add STS-2014

* add STS-2014

* add trainers and evaluators for Quora

* add quora in datasets/

* process the  merge confict in common/dataset.py

* add sts2014_evaluator.py and quora_evaluator.py
 2018-10-09 09:10:37 -04:00
Linqing Liu de56206fe4 add SNLI / STS-2014 / Quora dataset (#148) 
* add SNLI dataset

* add STS-2014

* add trainers and evaluators for Quora

* add quora in datasets/

* process the  merge confict in common/dataset.py
 2018-10-08 19:23:22 -04:00
Ashutosh-Adhikari 999e0c88f2 Add Reuters dataset option for common.dataset (#149) 
* Add Reuters option in common.dataset
 2018-10-07 00:57:38 -04:00
Achyudh Ram 1b817d3e24 WIP: Add Reuters-21578 dataset (#147) 
* Add ReutersTrainer, ReutersEvaluator options in Factory classes

* Add Reuters to Kim-CNN command line arguments

* Fix SST dataset path according to changes in Kim-CNN args

The dataset path in args.py was made to point at the dataset folder rather than dataset/SST folder. Hence SST folder was added to paths in the SST dataset class

* Add Reuters dataset class, and support in __main__

* Add Reuters dataset trainers and evaluators

* Remove debug print statement in reuters_evaluator

* Fix rounding bug in reuters_trainer and reuters_evaluator
 2018-10-02 21:09:51 -04:00
Victor Yang 10e4e56817 Add twitter url dataset (#145) 
* add prediction/qrel files dump option

* fix comment

* add twitter-url dataset, minor refactor

* fix minor error
 2018-09-08 11:16:23 +08:00
Victor Yang 5afb845a4b Add PIT2015 dataset (#141) 
* add prediction/qrel files dump option

* fix comment

* upgrade to torchtext 0.3

* remove ngram

* update device

* minor update

* add pit2015

* revert update torchtext

* revert update torchtext

* revert update torchtext
 2018-08-12 18:56:08 -04:00
Michael Tu 8a00f9cdcd Make Kim CNN ONNX-exportable (#136) 
* Kim CNN - only set embedding for corresponding mode

* Kim CNN ONNX Export

* Specify dummy ONNX input size from command line
 2018-08-04 17:30:24 -04:00
Michael Tu 82bf90f4bb Tune Kim CNN for SST-2 and Improve SST-1 Results with Dataset/Initialization Changes (#133) 
* SST change min_freq and Kim CNN init distribution

* Add tuned results for SST-1 and SST-2

* Fix typo
 2018-07-11 22:08:26 -04:00
Ralph Tang 5ff980d120 Add Apache 2.0 license (#132) 2018-07-10 22:06:55 -04:00
Michael Tu 8563ad5976 Kim CNN OOP Refactoring (#124) 
* Nuke obsolete artifacts

* Refactor Kim CNN

* Make kim_cnn a module

* Fix bugs

* Update README

* Add choices to dataset arg

* update for sst2

update sst.py

update sst.py

* Add Kim CNN dataset choices to args.py

* Update tuned SST-1 accuracy
 2018-07-03 16:42:31 -04:00
Linqing Liu fae229eba4 Update VDPWI README with new results (#131) 
- Update with results on SICK, TrecQA, and WikiQA datasets
 2018-07-03 09:37:18 -04:00
Linqing Liu a4e847725c Change default hyperparameters for VDPWI (#130) 
- Default hyperparameters on SICK now use different batch size, RNN hidden dimension, learning rate, optimizer, and number of epochs.
 2018-06-27 15:00:42 -04:00
Ralph Tang e4149ff400 Use PackedSequence for ConvRNN (#126) 
* Use PackedSequence in ConvRNNs instead
- Zero-padding is incompatible with the current model

* Update test script

* Clean up comments and logic

* Add TQDM to requirements

* Add results using default hyperparameters
 2018-06-24 09:48:16 -04:00
Michael Tu 2f7731b4d5 Update data path (#127) 2018-06-23 23:55:42 -04:00
Victor Yang 921a45e7ca add storing prediction/qrel files option for test set (#125) 
* add prediction/qrel files dump option

* fix comment
 2018-06-18 20:43:40 -04:00
Michael Tu 5cc027f03e Create Castor-level SST dataset (#123) 
* Move SST to root directory datasets

* Fix bugs
 2018-06-09 17:32:32 -04:00
Ralph Tang cb9fad9f97 Make ConvRNN compatible with PyTorch 0.4 (#122) 2018-06-07 20:40:40 -04:00
Jimmy Lin 90dc6af78c Update README.md (#120) 2018-05-27 09:19:58 -04:00
Jimmy Lin e9c011b421 Update README.md (#119) 2018-05-27 09:18:02 -04:00
Jimmy Lin bc74ae6531 Update README.md 2018-05-27 07:07:35 -04:00
Jimmy Lin c59739488c Update README.md 2018-05-27 07:06:32 -04:00
Jimmy Lin 97a76495fd Update requirements.txt (#116) 2018-05-26 22:53:26 -04:00
Jimmy Lin 3b426c1c25 Update README.md (#117) 2018-05-26 22:05:54 -04:00
Jimmy Lin bc58fc7607 Update README.md (#115) 2018-05-26 18:37:24 -04:00
Michael Tu 5bf33bf8ea Update README to use Castor-models and Instructions for Internal Users (#113) 
* Update instructions to use Castor-models
* Consolidate requirements.txt
* Refine README with convenience scripts
* Update internal instructions
* MP-CNN working dir minor edit
 2018-05-25 18:14:10 -04:00
Ralph Tang 62f8abec7f Add README (#110) 
* Make *QA/MSRVID work with VDPWI

* Add README
 2018-05-25 11:52:25 -04:00
Michael Tu ef05240819 Update MP-CNN Doc with Pre-Trained Models (#111) 
* Delete outdated troubleshooting section

* QAEvalutor bugfix

* Add instructions for pre-trained models
 2018-05-25 11:51:19 -04:00
Ralph Tang 494ce36575 Refactor VDPWI to use common API (#109) 2018-05-25 01:30:53 -04:00
Michael Tu 0c3a91c443 Check in MP-CNN Lite Model (#108) 
* Add MP-CNN Lite model

* MP-CNN Lite bug fixes
 2018-05-25 00:15:50 -04:00
Michael Tu d7a631b0a9 MP-CNN with Bugs Fixed and PyTorch v0.4 (#107) 
* Refactor datasets

* Update evaluators

* Update trainers

* Update main and MP-CNN model

* Add serialization util

* Fix bugs

* Refactoring for NCE to use new parent class
 2018-05-24 23:42:13 -04:00