Commits · 34ccc8ebf4c9e440e4621dab6cdb0a8f2dc903ad · chenpangpang / transformers

11 Apr, 2019 1 commit
- finetuning any BERT model - fixes #455 · 4bc4c69a
  thomwolf authored Apr 11, 2019
  
  4bc4c69a
25 Mar, 2019 1 commit
- Corrected the displayed loss when gradient_accumulation_steps > 1 · 8d1d1ffd
  Matthew Carrigan authored Mar 25, 2019
  
  8d1d1ffd
21 Mar, 2019 4 commits
- Added a --reduce_memory option to the training script to keep training · 7d1ae644
  Matthew Carrigan authored Mar 21, 2019
```
data on disc as a memmap rather than in memory
```
  7d1ae644
- Fixed up the notes on a possible future low-memory path · 8a861048
  Matthew Carrigan authored Mar 21, 2019
  
  8a861048
- Reduced memory usage for pregenerating the data a lot by writing it · a8a577ba
  Matthew Carrigan authored Mar 21, 2019
```
out on the fly without shuffling - the Sampler in the finetuning script
will shuffle for us.
```
  a8a577ba
- Reduced memory usage for pregenerating the data a lot by writing it · 0ae59e66
  Matthew Carrigan authored Mar 21, 2019
```
out on the fly without shuffling - the Sampler in the finetuning script
will shuffle for us.
```
  0ae59e66
20 Mar, 2019 5 commits
- PEP8 and formatting cleanups · 7de5c6aa
  Matthew Carrigan authored Mar 20, 2019
  
  7de5c6aa
- Added final TODOs · 1798e98e
  Matthew Carrigan authored Mar 20, 2019
  
  1798e98e
- Fixed embarrassing indentation problem · c64c2fc4
  Matthew Carrigan authored Mar 20, 2019
  
  c64c2fc4
- Fixed logging · 0540d360
  Matthew Carrigan authored Mar 20, 2019
  
  0540d360
- First commit of the new LM finetuning · 976554a4
  Matthew Carrigan authored Mar 20, 2019
  
  976554a4