Skip to content

Conversation

@seemethere
Copy link
Member

@seemethere seemethere commented Jun 15, 2020

Model was not getting initialized property since it was using the
decoder object instead of decoder weight to initialize zeros.

Closes #783

Solution can be attributed to @mjoukamaa

Signed-off-by: Eli Uriegas [email protected]

Model was not getting initialized property since it was using the
decoder object instead of decoder weight to initialize zeros.

Signed-off-by: Eli Uriegas <[email protected]>
@seemethere seemethere requested review from soumith and vishwakftw June 15, 2020 22:17
Copy link
Contributor

@vishwakftw vishwakftw left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looks good to me, thanks.

The blame is attributed to me due to #767.

@seemethere seemethere merged commit 59caa16 into pytorch:master Jun 15, 2020
@seemethere seemethere deleted the fix_transformer branch June 15, 2020 22:47
YinZhengxun pushed a commit to YinZhengxun/mt-exercise-02 that referenced this pull request Mar 30, 2025
word_language_model: Fix Transformer init_weights
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Bug in TransformerModel for word_language_model initialization

2 participants