GitLab

This reverts commit b2e119c209363e6ff6d2878a69c7d1a507a2e9be.

Improve dataloader speed and deprecate concept of batch_offset (use --sample-without-replacement instead)

- process only last time step during generation
- cache keys and values
- dont apply masking during generation

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

Update training commands

specify a single GPU setup for IWSLT14

Update training commands in data/README to match the latest version of this project according to #132.

Continue from 3c072958: add omitted "\".

Update training commands in data/README to match the latest version of this project according to #132.

- Motivation: in the previous data/README, the commands are obsolete and will cause the error "unrecognized arguments: --label-smoothing 0.1 --force-anneal 50". 
- What's changed: add arguments "--criterion label_smoothed_cross_entropy" and "--lr-scheduler fixed" to the training commands of all 3 datasets.
- Result: the new commands run without error on all 3 datasets.