Changes in TransformerDecoder break arch compat with published models #2664
Comments
|
There should be an automatic upgrade path for old checkpoints: https://github.com/pytorch/fairseq/blob/fd080b308e1e3361d6c498b235496080fa6599e5/fairseq/models/transformer.py#L848-L858 Please let us know if this doesn't work! We also have some internal tests that should confirm that all of the released models run, although again please let us know if something doesn't and we can add more tests. |
Closed
facebook-github-bot
pushed a commit
that referenced
this issue
Nov 20, 2021
Summary: Fix a typo # Before submitting - [ ] Was this discussed/approved via a Github issue? (no need for typos, doc improvements) - [ ] Did you read the [contributor guideline](https://github.com/pytorch/fairseq/blob/main/CONTRIBUTING.md)? - [ ] Did you make sure to update the docs? - [ ] Did you write any new necessary tests? ## What does this PR do? Fixes # (issue). ## PR review Anyone in the community is free to review the PR once the tests have passed. If we didn't discuss your PR in Github issues there's a high chance it will not be merged. ## Did you have fun? Make sure you had fun coding � Pull Request resolved: fairinternal/fairseq-py#2664 Reviewed By: alexeib Differential Revision: D32577174 Pulled By: arbabu123 fbshipit-source-id: 9b65619843fd9684be319024ccc9e6b0681db7de
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Example:
transformer_vaswani_wmt_en_de_big.Specifically,
output_projectionin the case thatshare_input_output_embed == True.Model example: https://dl.fbaipublicfiles.com/fairseq/models/wmt14.en-fr.joined-dict.transformer.tar.bz2, from: https://github.com/pytorch/fairseq/tree/master/examples/translation
v0.9.0:
https://github.com/pytorch/fairseq/blob/df2f84ce619edddb80e720a56abc74d5490fed99/fairseq/models/transformer.py#L538-L551
master:
https://github.com/pytorch/fairseq/blob/fd080b308e1e3361d6c498b235496080fa6599e5/fairseq/models/transformer.py#L620-L633
The text was updated successfully, but these errors were encountered: