aayush
03c4a2cd15
Fixed self attention, local tests passed
2019-02-15 02:10:18 +05:30
Aayush
8fe2891402
Make callable through bicls_head
2019-01-07 00:02:39 +05:30
Aayush
d0c15472d6
Self attention for pooling linear classifier
...
This PR will introduce a `BiAttentionPoolingClassifier` as in [Attention is all you need](https://arxiv.org/abs/1706.03762 ) following the discussion with @sebastianruder in Teams.
I ran out of memory on my 1060 while testing the attention module, but was able to at least verify that it is functionally correct. Some changes might be required to ensure that the tensor passed to `self.layers` is of the right shape (but I'm not quite sure as of now).
I'll shift all the stuff to Collab for testing and see if it's any help.
2019-01-06 23:58:09 +05:30
Piotr Czapla
0085c18ae0
Fix splitting by article for local languages and make it more memory efficient
...
The issue was that the code assumed that empty lines have space followed by a new line, which isn't the case for datasets generated by our scripts.
2019-01-03 12:16:41 +01:00
Piotr Czapla
8a2fed41c5
Merge pull request #18 from n-waves/refactor
...
[WIP] Refactoring #17
2019-01-01 15:13:57 +01:00
Piotr Czapla
f784d7bcd2
Make the article detection code work with our wikitext
2019-01-01 15:13:14 +01:00
Piotr Czapla
0945c699c7
Fixes #25 by adding article title in markdown format to wiki text
2019-01-01 15:07:02 +01:00
Piotr Czapla
4a858f3572
Simplfy and unify input data parsing
2019-01-01 14:43:12 +01:00
Piotr Czapla
d269d53d7d
Use fastai tokens instead of <xxx>
...
f'xx{token_name}' are kept as one token by Moses tokenizer, which is sometimes required if you want to have moses in tokenizers pipeline, and we use that for imdb.
2019-01-01 14:40:59 +01:00
Piotr Czapla
514a9e6b86
Fix BiLM training after update to newest fastai
2018-12-31 12:13:53 +01:00
Piotr Czapla
f1b49a0d34
fix imports in learner (fastai adpatation)
2018-12-27 15:15:09 +01:00
Piotr Czapla
82d6a30b11
Fix SentencePiece implementation, to get 94.5% on imdb
...
- train on whole articles when tokenizer is sentencepice
- add moses tokenizer to get the same tokens on imdb as on wt103
- apply pre / post processing rules to moses tokenzier so that sentencepiece works on correctly preprocessed text (lowercased with html removed)
- add alpha implementation on xnli (no tests)
- remove vocab adaptation when swiching from wiki to imdb. As otherwise we get 50% of missing words and 93% accuracy on imdb
2018-12-27 15:14:41 +01:00
Piotr Czapla
ff30fb5642
Fastai upgrade
2018-12-27 14:57:15 +01:00
Piotr Czapla
dca502a398
Add docs how to train a classfier
2018-12-12 00:31:48 +01:00
Piotr Czapla
f9394b9af1
Add callbacks to save history and best weights remove bs & drop_mult
2018-12-12 00:31:16 +01:00
Piotr Czapla
fac3ce343b
Fix BiLM implementation after upgrade of fastai
2018-12-12 00:29:23 +01:00
Piotr Czapla
1b9d04d7ef
Expose use_test_for_validation as param to train
2018-12-09 22:45:21 +01:00
Piotr Czapla
34b1d18600
load wikipedia as articles for v & fv tok.
2018-12-09 22:42:17 +01:00
Piotr Czapla
2338618563
Make biclasifier head a hyperparameter.
2018-12-09 00:39:27 +01:00
Piotr Czapla
591393b96c
Change AvgPooling to BiPooling as a default
2018-12-08 23:48:02 +01:00
Piotr Czapla
978bbcc923
Fix issue with finetuning language model (it wasn't freezed)
2018-12-08 23:10:31 +01:00
Piotr Czapla
0b7dc7ec6b
Add two experiments showing how to train imdb classification to 94% accuracy
2018-12-07 16:10:32 +01:00
Piotr Czapla
50ab34eea9
Fix classification scripts and give a way to test accuracy on test set
2018-12-07 16:10:20 +01:00
Piotr Czapla
4ae2a158a2
Respect batch size in training lm model.
2018-12-06 23:29:43 +01:00
Piotr Czapla
6a899015ed
Generate imdb unsp.csv
2018-12-05 16:27:41 +01:00
Piotr Czapla
2acbf15555
Tweak hyper training params of cls (drop_mul, bs, true_wd=True)
...
I've set the same hyperparams as in lesson3
2018-12-05 16:27:14 +01:00
Piotr Czapla
4454af167f
Use more text during pretraining of imdb
2018-12-05 16:26:16 +01:00
Piotr Czapla
5524006d81
Respect max_vocab
2018-12-04 16:41:22 +01:00
Piotr Czapla
908c3d7e8a
bug fix
2018-12-04 16:41:07 +01:00
Piotr Czapla
f25deb3049
Use trn + tst for LM training
2018-12-04 01:43:01 +01:00
Piotr Czapla
9826f6881c
Fix the way trn & val set is created in imdb
2018-12-04 01:35:59 +01:00
Piotr Czapla
039624870a
Fix loading tokenized data set in train cls
2018-12-04 01:28:03 +01:00
Piotr Czapla
a499bf9a20
Make cls train work with relative paths
2018-12-04 01:24:44 +01:00
Piotr Czapla
35a5dcb75a
Fix issue when running cls training from command line
2018-12-04 00:59:36 +01:00
Piotr Czapla
82c955ce6a
Add different tokenization algorithms to train_clas
2018-12-04 00:51:21 +01:00
Piotr Czapla
83427aadc6
Add moses with fastai preprocessing
2018-12-02 21:43:57 +01:00
Piotr Czapla
b9eb7388f6
Fix bidir for fastai tokenizer
2018-12-01 23:58:14 +01:00
Piotr Czapla
0c4aed6d05
Really fix conversion from str to Tokenzier
2018-12-01 16:51:21 +01:00
Piotr Czapla
e14c967cc8
Fix string parsing in tokenzier
2018-12-01 16:46:25 +01:00
Piotr Czapla
887211137a
Add fastai tokenizer to pretrain_lm
2018-12-01 16:42:11 +01:00
Piotr Czapla
c17dcce75e
spelling
2018-12-01 15:24:30 +01:00
Piotr Czapla
ab9faa2ad9
Clean up Fire interface.
2018-12-01 13:42:10 +01:00
Piotr Czapla
4b29376b44
Rewrite classifier to use changed pretrain_lm
2018-12-01 10:58:46 +01:00
Piotr Czapla
b3f5ae1ad5
Clean the way we save models
2018-12-01 10:58:23 +01:00
Piotr Czapla
6d6ebef1ca
Update to newst fastai
2018-12-01 10:57:20 +01:00
Piotr Czapla
80d4d4da29
Extract params to an experiment data class
...
You can run this as follows:
`python -m ulmfit.pretrain_lm --dir-path 'data/wiki/wikitext-2' --qrnn=True train_lm --num_epochs=1`
2018-11-25 12:37:45 +01:00
Piotr Czapla
eea9be09db
Merge pull request #6 from n-waves/bilm
...
Bidirectional language model + fixes to the end-to-end tests
2018-11-24 23:58:00 +01:00
Piotr Czapla
be117abac4
Make the end to end test run correctly
2018-11-24 23:51:23 +01:00
Piotr Czapla
8da47324c2
Clean ups and fixes
2018-11-22 15:32:40 +01:00
Piotr Czapla
dbd4884228
Fixes after mergin with master and updateing to newset fastai
2018-11-22 01:14:38 +01:00
aayush