Skip to content
This repository has been archived by the owner on Nov 3, 2023. It is now read-only.

2020-07-16
Compare
Choose a tag to compare
@stephenroller stephenroller released this 16 Jul 20:39
v0.1.20200716
1a65ad0
This commit was created on GitHub.com and signed with GitHub’s verified signature. The key has expired.
GPG key ID: 4AEE18F83AFDEB23
Expired
Learn about vigilant mode.

Notable additions since 2020-07-13:

  • Support for adding special to tokenizers (#2828)
  • Distributed Evaluation, to help scale generations/evaluations across many GPUs (#2775)
  • Google SGD dataset (#2738).

Other notable features since 2020-06-16:

  • Multi-turn chunk teacher (#2805)
  • Fromfile jsonl teacher; works directly from files dumped by self_chat, ACUTE logs, etc. (#2717)
  • BART model now part of parlai (-m bart) (#2781)
  • Atomic saves to be more robust to preemption (#2760)

Recent bugfixes:

  • #2789: crash when printing an empty report
  • #2788: multi-tasking failed to sample appropriately with --datatype train:stream.
Assets 2
Loading