Commits · 4fa8760e9a1f93320007aa9060e7c60fd2e90cc7 · OpenDAS / Fairseq

15 Jun, 2018 2 commits
- Improve dataloader speed and deprecate concept of batch_offset (use... · 4fa8760e
  Myle Ott authored Apr 07, 2018
```
Improve dataloader speed and deprecate concept of batch_offset (use --sample-without-replacement instead)
```
  4fa8760e
- Simulated big batches · 2d27ae08
  Sergey Edunov authored Apr 07, 2018
  
  2d27ae08
02 Apr, 2018 1 commit

Merge internal changes (#136) · d3795d6c

Myle Ott authored Apr 02, 2018

Changes:
- 7d19e36: Add `--sampling` flag to generate.py to sample instead of doing beam search
- c777340: Add `scripts/average_checkpoints.py` to average multiple checkpoints into a combined model
- 3ea882c: Add `--max-update` option to train.py to stop training after a given number of updates
- small bugfixes for distributed training, LSTM, inverse square root LR scheduler

d3795d6c

27 Feb, 2018 2 commits

Add OOM counter back to logging output · b9f2d427
Myle Ott authored Feb 14, 2018

b9f2d427

fairseq-py goes distributed (#106) · 66415206

Myle Ott authored Feb 27, 2018

This PR includes breaking API changes to modularize fairseq-py and adds support for distributed training across multiple nodes.

Changes:
- c7033ef: add support for distributed training! See updated README for usage.
- e016299: modularize fairseq-py, adding support for register_model, register_criterion, register_optimizer, etc.
- 154e440: update LSTM implementation to use PackedSequence objects in the encoder, better following best practices and improving perf
- 90c2973 and 1da6265: improve unit test coverage

66415206