Commit Graph

Select branches

Hide Pull Requests

BCELoss

bilm

cls

datasets

distill

lbl_smoothing

master

poleval2018

polyglot-lm

pretrain-lm

reproduce-poleval

self-attention-module

text_cols

thai2fit

tmp

ulmfit-multilingual-original-scripts

xnli

Mono Color

• 593c5661bf Added alpha and beta params for RNNTrainer Tomasz Pietruszka 2019-01-13 17:26:16 +01:00
• e6407fe0c9 Added the param and model type for BwdLM Tomasz Pietruszka 2019-01-10 00:45:22 +01:00
• 8fe2891402 Make callable through bicls_head Aayush 2019-01-07 00:02:39 +05:30
• d0c15472d6 Self attention for pooling linear classifier Aayush 2019-01-06 23:58:09 +05:30
• 0e2e058b82 Simplify finetuning using new helper methods NAUSICAA\Julian 2019-01-03 17:02:56 -03:00
• 39d38f0f46 Restore newline NAUSICAA\Julian 2019-01-03 09:40:31 -03:00
• 0856d33fa8 Merge issues NAUSICAA\Julian 2019-01-03 09:35:17 -03:00
• 3f91460dc8 Merge branch 'master' of https://github.com/n-waves/ulmfit-multilingual into polyglot-lm NAUSICAA\Julian 2019-01-03 09:30:54 -03:00
• 0085c18ae0 Fix splitting by article for local languages and make it more memory efficient Piotr Czapla 2019-01-03 12:16:41 +01:00
• 52702e32ae Merge branch 'master' of https://github.com/n-waves/ulmfit-multilingual into polyglot-lm NAUSICAA\Julian 2019-01-02 18:41:07 -03:00
• 8a2fed41c5 Merge pull request #18 from n-waves/refactor Piotr Czapla 2019-01-01 15:13:57 +01:00
• f784d7bcd2 Make the article detection code work with our wikitext Piotr Czapla 2019-01-01 15:13:14 +01:00
• 0945c699c7 Fixes #25 by adding article title in markdown format to wiki text Piotr Czapla 2019-01-01 15:07:02 +01:00
• 4a858f3572 Simplfy and unify input data parsing Piotr Czapla 2019-01-01 14:43:12 +01:00
• d269d53d7d Use fastai tokens instead of <xxx> Piotr Czapla 2019-01-01 14:40:59 +01:00
• 9d529b1fb6 Merge branch 'refactor' of https://github.com/n-waves/ulmfit-multilingual into polyglot-lm NAUSICAA\Julian 2018-12-31 11:35:25 -03:00
• 514a9e6b86 Fix BiLM training after update to newest fastai Piotr Czapla 2018-12-31 12:13:53 +01:00
• 6780559b46 Dataset name NAUSICAA\Julian 2018-12-29 17:13:38 -03:00
• b78a2ebaad Fixing typo NAUSICAA\Julian 2018-12-29 17:12:38 -03:00
• 30565a2148 Remove bad imports Julian Eisenschlos 2018-12-29 20:09:46 +00:00
• d439f814d3 Validate other languages NAUSICAA\Julian 2018-12-29 17:02:14 -03:00
• f1b49a0d34 fix imports in learner (fastai adpatation) Piotr Czapla 2018-12-27 15:15:09 +01:00
• 82d6a30b11 Fix SentencePiece implementation, to get 94.5% on imdb Piotr Czapla 2018-12-27 15:14:41 +01:00
• ff30fb5642 Fastai upgrade Piotr Czapla 2018-12-27 14:57:15 +01:00
• 15b49fc4f3 Merge branch 'refactor' of https://github.com/n-waves/ulmfit-multilingual into polyglot-lm NAUSICAA\Julian 2018-12-24 17:17:28 -03:00
• 0436ba7ca1 Rename train_xlingual_cls.py to train_xlingual_clas.py Julian Eisenschlos 2018-12-24 17:09:22 -03:00
• 3ef3be1b93 Create train_xlingual_cls.py Julian Eisenschlos 2018-12-24 17:02:37 -03:00
• a8f51900bc Quiet wget datasets NAUSICAA\Julian 2018-12-19 08:40:25 -03:00
• 08102657bb Remove verbose flag NAUSICAA\Julian 2018-12-19 08:18:31 -03:00
• dca502a398 Add docs how to train a classfier Piotr Czapla 2018-12-12 00:31:48 +01:00
• f9394b9af1 Add callbacks to save history and best weights remove bs & drop_mult Piotr Czapla 2018-12-12 00:31:16 +01:00
• fac3ce343b Fix BiLM implementation after upgrade of fastai Piotr Czapla 2018-12-12 00:29:23 +01:00
• 1b9d04d7ef Expose use_test_for_validation as param to train Piotr Czapla 2018-12-09 22:45:21 +01:00
• 34b1d18600 load wikipedia as articles for v & fv tok. Piotr Czapla 2018-12-09 22:42:17 +01:00
• 2338618563 Make biclasifier head a hyperparameter. Piotr Czapla 2018-12-09 00:39:27 +01:00
• 591393b96c Change AvgPooling to BiPooling as a default Piotr Czapla 2018-12-08 23:48:02 +01:00
• 978bbcc923 Fix issue with finetuning language model (it wasn't freezed) Piotr Czapla 2018-12-08 23:10:31 +01:00
• 0b7dc7ec6b Add two experiments showing how to train imdb classification to 94% accuracy Piotr Czapla 2018-12-07 16:10:32 +01:00
• 50ab34eea9 Fix classification scripts and give a way to test accuracy on test set Piotr Czapla 2018-12-07 16:10:20 +01:00
• 4ae2a158a2 Respect batch size in training lm model. Piotr Czapla 2018-12-06 23:29:43 +01:00
• 6a899015ed Generate imdb unsp.csv Piotr Czapla 2018-12-05 16:27:41 +01:00
• 2acbf15555 Tweak hyper training params of cls (drop_mul, bs, true_wd=True) Piotr Czapla 2018-12-05 16:27:14 +01:00
• 4454af167f Use more text during pretraining of imdb Piotr Czapla 2018-12-05 16:26:16 +01:00
• 5524006d81 Respect max_vocab Piotr Czapla 2018-12-04 16:41:22 +01:00
• 908c3d7e8a bug fix Piotr Czapla 2018-12-04 16:41:07 +01:00
• f25deb3049 Use trn + tst for LM training Piotr Czapla 2018-12-04 01:43:01 +01:00
• 9826f6881c Fix the way trn & val set is created in imdb Piotr Czapla 2018-12-04 01:35:59 +01:00
• 039624870a Fix loading tokenized data set in train cls Piotr Czapla 2018-12-04 01:28:03 +01:00
• a499bf9a20 Make cls train work with relative paths Piotr Czapla 2018-12-04 01:24:44 +01:00
• 35a5dcb75a Fix issue when running cls training from command line Piotr Czapla 2018-12-04 00:59:36 +01:00
• 82c955ce6a Add different tokenization algorithms to train_clas Piotr Czapla 2018-12-04 00:51:21 +01:00
• 83427aadc6 Add moses with fastai preprocessing Piotr Czapla 2018-12-02 21:43:57 +01:00
• b9eb7388f6 Fix bidir for fastai tokenizer Piotr Czapla 2018-12-01 23:58:14 +01:00
• 0c4aed6d05 Really fix conversion from str to Tokenzier Piotr Czapla 2018-12-01 16:51:21 +01:00
• e14c967cc8 Fix string parsing in tokenzier Piotr Czapla 2018-12-01 16:46:25 +01:00
• 887211137a Add fastai tokenizer to pretrain_lm Piotr Czapla 2018-12-01 16:42:11 +01:00
• c17dcce75e spelling Piotr Czapla 2018-12-01 15:24:30 +01:00
• ab9faa2ad9 Clean up Fire interface. Piotr Czapla 2018-12-01 13:42:10 +01:00
• 4b29376b44 Rewrite classifier to use changed pretrain_lm Piotr Czapla 2018-12-01 10:58:46 +01:00
• b3f5ae1ad5 Clean the way we save models Piotr Czapla 2018-12-01 10:58:23 +01:00
• 6d6ebef1ca Update to newst fastai Piotr Czapla 2018-12-01 10:57:20 +01:00
• bf33f222da callback > callback_fns NAUSICAA\Julian 2018-11-26 22:48:42 -03:00
• 4d4ef1ae95 Change callback_fns NAUSICAA\Julian 2018-11-26 22:43:39 -03:00
• 8074041965 Cache intermediate results NAUSICAA\Julian 2018-11-26 21:24:19 -03:00
• 699093725f Script to merge two language token files NAUSICAA\Julian 2018-11-26 14:49:38 -03:00
• c4e0b7b3d6 Lower memory footprint NAUSICAA\Julian 2018-11-25 23:47:24 -03:00
• 80d4d4da29 Extract params to an experiment data class Piotr Czapla 2018-11-25 12:37:45 +01:00
• eea9be09db Merge pull request #6 from n-waves/bilm Piotr Czapla 2018-11-24 23:58:00 +01:00
• be117abac4 Make the end to end test run correctly bilm Piotr Czapla 2018-11-24 23:51:23 +01:00
• 8da47324c2 Clean ups and fixes Piotr Czapla 2018-11-22 15:32:40 +01:00
• dbd4884228 Fixes after mergin with master and updateing to newset fastai Piotr Czapla 2018-11-22 01:14:38 +01:00
• 934fc79384 Merge branch 'master' into bilm Piotr Czapla 2018-11-21 23:48:23 +01:00
• 9aa877dcd0 Share trained LM between different classfiication runs Piotr Czapla 2018-11-21 18:51:42 +01:00
• 979eb196d8 Change the classfication training learning rate to the one that was working te best in my exp. on bidirectional clasification Piotr Czapla 2018-11-21 18:46:05 +01:00
• 6e3ef21b1f Add Avg BiClassifier Piotr Czapla 2018-11-21 18:44:49 +01:00
• cbed02d5e0 Merge branch 'master' into bilm Piotr Czapla 2018-11-21 18:23:54 +01:00
• 8cb867b066 Compatibility with new fastai version NAUSICAA\Julian 2018-11-20 19:21:02 -03:00
• 79691791b3 Typo fix NAUSICAA\Julian 2018-11-20 18:53:23 -03:00
• 88d94f38ba Merge pull request #16 from n-waves/sentencepiece_fixes Julian Eisenschlos 2018-11-20 17:14:41 -03:00
• 40a2990322 Add suggested changes to rules system NAUSICAA\Julian 2018-11-20 09:06:03 -03:00
• a36c0518c6 Merge branch 'master' of https://github.com/n-waves/ulmfit-multilingual into sentencepiece_fixes NAUSICAA\Julian 2018-11-20 08:55:37 -03:00
• 429a3fe4c2 Merge pull request #13 from n-waves/models_path_fix Aayush 2018-11-20 12:25:25 +05:30
• 3b5629c0c5 Removing files NAUSICAA\Julian 2018-11-20 00:46:04 -03:00
• 0ecc6342a6 Fixes when running sentence piece end to end NAUSICAA\Julian 2018-11-20 00:38:44 -03:00
• 2a36762837 Merge branch 'master' of https://github.com/n-waves/ulmfit-multilingual into datasets NAUSICAA\Julian 2018-11-19 16:34:15 -03:00
• a08881180c Merge branch 'master' of https://github.com/n-waves/ulmfit-multilingual into polyglot-lm NAUSICAA\Julian 2018-11-19 16:33:44 -03:00
• 470f984ca4 Script to merge two languages NAUSICAA\Julian 2018-11-19 16:31:38 -03:00
• 2bde30a402 Add experiment with biclassifier (concat) Piotr Czapla 2018-11-19 15:31:10 +01:00
• 9f526a9849 Add experiment where wd_true is set to True. Piotr Czapla 2018-11-19 15:30:28 +01:00
• 73403bdcb3 fix double EOS in read_imdb Aayush 2018-11-19 19:30:11 +05:30
• 6e867811db Support for running sentencepiece with train_clas Aayush 2018-11-19 19:28:19 +05:30
• 5ffbe8ba5c add vocab_size to sentencepiece Aayush 2018-11-19 19:13:55 +05:30
• 69ed7b4169 Added preliminary test scripts for sentencepiece Aayush 2018-11-19 19:11:20 +05:30
• 7f1f8efcc3 Working version of biclassfier Piotr Czapla 2018-11-19 12:58:58 +01:00
• c821d2e783 first version of bi classifier Piotr Czapla 2018-11-19 09:59:08 +01:00
• 56a9feec7d Removed dependency note, use requirements.txt Nirant 2018-11-19 12:22:41 +05:30
• 635fb1c113 Add prepare_xnli.py Nirant K 2018-11-19 06:50:50 +00:00
• 6292412eff Adding scripts to download and format RCV and Webis CLS X-Language datasets NAUSICAA\Julian 2018-11-18 22:29:07 -03:00
• 37b73e262f Fix bug where de-all was de-100 Piotr Czapla 2018-11-17 17:14:13 +01:00
• 2674a713fc fix resuming training of classifier Piotr Czapla 2018-11-17 17:13:15 +01:00