GitLab

Summary:
Changelog:
- `90f52a1`: Support loading subsets of the data on each worker with the `--fix-batches-to-gpus` flag. This should fix #217 and #266.
- `6eda0a9`: Update README for replicating the "Scaling Neural Machine Translation" paper
- `b14c7cf`: Fallback to no_c10d backend for pytorch 0.4.1 (fixes #294)
Pull Request resolved: https://github.com/pytorch/fairseq/pull/295

Differential Revision: D10121559

Pulled By: myleott

fbshipit-source-id: 41c84d0ee4cdd113544b5d3aa38ae8b23acc2c27

Co-authored-by: liezl200 <lie@fb.com>

- no more FP16Trainer, we just have an FP16Optimizer wrapper
- most of the distributed code is moved to a new wrapper class called DistributedFairseqModel, which behaves like DistributedDataParallel and a FairseqModel at the same time
- Trainer now requires an extra dummy_batch argument at initialization, which we do fwd/bwd on when there's an uneven number of batches per worker. We hide the gradients from these dummy batches by multiplying the loss by 0
- Trainer.train_step now takes a list of samples, which will allow cleaner --update-freq

Update readme with WMT'18 model (#433)

Oss master

also don't crash if param does not recieve grads