Commits · 3951c2c189a73ac3bdcef1c925313cd0be576239 · chenpangpang / transformers

27 Nov, 2018 1 commit
- Bug fix in examples;correct t_total for distributed training;run prediction for full dataset · 0aaedcc0
  Li Li authored Nov 27, 2018
  
  0aaedcc0
26 Nov, 2018 2 commits
- remove convert_to_unicode and printable_text from examples · 32167cdf
  thomwolf authored Nov 26, 2018
  
  32167cdf
- update cache_dir in readme and examples · 05053d16
  thomwolf authored Nov 26, 2018
  
  05053d16
20 Nov, 2018 1 commit
- fixing param.grad is None in fp16 examples · 2f21497d
  thomwolf authored Nov 20, 2018
  
  2f21497d
18 Nov, 2018 1 commit
- Fix ineffective no_decay bug · 6c4789e4
  xiaoda99 authored Nov 18, 2018
  
  6c4789e4
17 Nov, 2018 4 commits
- remove unused arguments in example scripts · aa50fd19
  thomwolf authored Nov 17, 2018
  
  aa50fd19
- clean up readme and examples · c8cba677
  thomwolf authored Nov 17, 2018
  
  c8cba677
- fix tests · 757750d6
  thomwolf authored Nov 17, 2018
  
  757750d6
- updating examples · 4e46affc
  thomwolf authored Nov 17, 2018
  
  4e46affc
15 Nov, 2018 1 commit
- preparing for first release · 1de35b62
  thomwolf authored Nov 15, 2018
  
  1de35b62
12 Nov, 2018 1 commit
- add fp16 training · 66b00908
  thomwolf authored Nov 12, 2018
  
  66b00908
10 Nov, 2018 1 commit
- fixing learning rate schedule when using gradient_accumulation_steps · a81a1ef8
  thomwolf authored Nov 10, 2018
  
  a81a1ef8
09 Nov, 2018 1 commit
- remove duplicate accumulate gradient step arguments · 34bdc8b5
  thomwolf authored Nov 09, 2018
  
  34bdc8b5
07 Nov, 2018 3 commits
- typo fix in output tuple · 48d4a531
  thomwolf authored Nov 07, 2018
  
  48d4a531
- cleaning up - speeding up a bit multi-gpu · dbc318a4
  thomwolf authored Nov 07, 2018
  
  dbc318a4
- fixing pre-processing bug - averaging loss for gradient accumulation - no_grad on evaluation · 6bb7510a
  thomwolf authored Nov 07, 2018
  
  6bb7510a
04 Nov, 2018 10 commits
- Fix `args.gradient_accumulation_steps` used before assigment. · 290633b8
  VictorSanh authored Nov 04, 2018
  
  290633b8
- Fix bug train_batch_size not an int. · 649e9774
  VictorSanh authored Nov 04, 2018
```
Division makes args.train_batch_size becoming a float.
cc @thomwolf
```
  649e9774
- Small logger bug (multi-gpu, distribution) in training · d55c3ae8
  VictorSanh authored Nov 04, 2018
  
  d55c3ae8
- clean up + mask is long · 3ddff783
  thomwolf authored Nov 04, 2018
  
  3ddff783
- remove old methods · c6207d85
  thomwolf authored Nov 04, 2018
  
  c6207d85
- add distributed training · 965b2565
  thomwolf authored Nov 04, 2018
  
  965b2565
- add gradient accumulation · 1ceac85e
  thomwolf authored Nov 04, 2018
  
  1ceac85e
- clean up · 6b0da96b
  thomwolf authored Nov 04, 2018
  
  6b0da96b
- logging + update copyright · 834b485b
  thomwolf authored Nov 04, 2018
  
  834b485b
- multi-gpu cleanup · 1701291e
  thomwolf authored Nov 04, 2018
  
  1701291e
03 Nov, 2018 13 commits
- update readme, file names, removing TF code, moving tests · f8276008
  thomwolf authored Nov 03, 2018
  
  f8276008
- fix typo · 1ba5b58c
  VictorSanh authored Nov 03, 2018
  
  1ba5b58c
- Fix both loss and eval metrics -> more coherence on the loss (eval vs train and tf vs pt) · 5858e8e4
  VictorSanh authored Nov 03, 2018
  
  5858e8e4
- Fix import on initalization · cd09cd5b
  VictorSanh authored Nov 03, 2018
  
  cd09cd5b
- special edition script · 04287a4d
  thomwolf authored Nov 03, 2018
  
  04287a4d
- Add seed in initialization · a1af5247
  VictorSanh authored Nov 03, 2018
  
  a1af5247
- Fix loss loss logging for multi-gpu compatibility · 4faeb38b
  Ubuntu authored Nov 03, 2018
  
  4faeb38b
- add model.zero_grad() · cb76c1dd
  thomwolf authored Nov 03, 2018
  
  cb76c1dd
- fixing optimization · 088ad458
  thomwolf authored Nov 03, 2018
  
  088ad458
- Create DataParallel model if several GPUs · 5f432480
  VictorSanh authored Nov 03, 2018
  
  5f432480
- Small typo in `trange` · 391a4ec2
  VictorSanh authored Nov 03, 2018
```
I seriously don't understand why they defined num_train_epochs as a float in the originial tf code.
I Will change it at the end to avoir merge conflicts for now.
```
  391a4ec2
- fix optimization_test · 0d8d2285
  thomwolf authored Nov 03, 2018
  
  0d8d2285
- removing f-string · 45efc9d8
  thomwolf authored Nov 03, 2018
  
  45efc9d8
02 Nov, 2018 1 commit
- Move command-line argparse arguments into main() function · 3ff2ec5e
  Tim Rault authored Nov 02, 2018
  
  3ff2ec5e