Thanks to visit codestin.com
Credit goes to github.com

Skip to content

Fix cuda native loss_ctc for varying input length#15798

Closed
t-vi wants to merge 1 commit into
pytorch:masterfrom
t-vi:fix_ctc_loss_varying_input_length
Closed

Fix cuda native loss_ctc for varying input length#15798
t-vi wants to merge 1 commit into
pytorch:masterfrom
t-vi:fix_ctc_loss_varying_input_length

Conversation

@t-vi
Copy link
Copy Markdown
Collaborator

@t-vi t-vi commented Jan 7, 2019

Thank you, @freesouls, for the reproducing example!

This is strictly fixing the bug in gradients for varying length inputs discussed in the middle-to-bottom of the bug report. I'll have a feature patch regarding inf losses -> NaN grads separately.

Fixes: #14401

@t-vi t-vi changed the title Fix loss_ctc for varying input length (fixes: #14401) Fix cuda native loss_ctc for varying input length Jan 7, 2019
@soumith soumith added this to the 1.0.1 milestone Jan 7, 2019
@t-vi
Copy link
Copy Markdown
Collaborator Author

t-vi commented Jan 7, 2019

So everything but MacOS seems to work and would just need a rebuild as the issue is fixed by @ssnl's #15795 (which just landed, thank you!).

@freesouls
Copy link
Copy Markdown

glad to hear that, hoping for the new release

Copy link
Copy Markdown
Contributor

@facebook-github-bot facebook-github-bot left a comment

Choose a reason for hiding this comment

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

@soumith is landing this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

zdevito pushed a commit to zdevito/ATen that referenced this pull request Jan 9, 2019
Summary:
Thank you, freesouls, for the reproducing example!

This is strictly fixing the bug in gradients for varying length inputs discussed in the middle-to-bottom of the bug report. I'll have a feature patch regarding inf losses -> NaN grads separately.

Fixes: #14401
Pull Request resolved: pytorch/pytorch#15798

Differential Revision: D13605739

Pulled By: soumith

fbshipit-source-id: 167ff42399c7e4cdfbd88d59bac5d25b57c0363f
@soumith soumith added the cherry-picked This PR was cherry-picked onto a release branch from master label Jan 17, 2019
soumith pushed a commit that referenced this pull request Jan 17, 2019
Summary:
Thank you, freesouls, for the reproducing example!

This is strictly fixing the bug in gradients for varying length inputs discussed in the middle-to-bottom of the bug report. I'll have a feature patch regarding inf losses -> NaN grads separately.

Fixes: #14401
Pull Request resolved: #15798

Differential Revision: D13605739

Pulled By: soumith

fbshipit-source-id: 167ff42399c7e4cdfbd88d59bac5d25b57c0363f
soumith pushed a commit that referenced this pull request Jan 18, 2019
Summary:
Thank you, freesouls, for the reproducing example!

This is strictly fixing the bug in gradients for varying length inputs discussed in the middle-to-bottom of the bug report. I'll have a feature patch regarding inf losses -> NaN grads separately.

Fixes: #14401
Pull Request resolved: #15798

Differential Revision: D13605739

Pulled By: soumith

fbshipit-source-id: 167ff42399c7e4cdfbd88d59bac5d25b57c0363f
laurentdupin pushed a commit to laurentdupin/pytorch that referenced this pull request Apr 24, 2026
Summary:
Thank you, freesouls, for the reproducing example!

This is strictly fixing the bug in gradients for varying length inputs discussed in the middle-to-bottom of the bug report. I'll have a feature patch regarding inf losses -> NaN grads separately.

Fixes: pytorch#14401
Pull Request resolved: pytorch#15798

Differential Revision: D13605739

Pulled By: soumith

fbshipit-source-id: 167ff42399c7e4cdfbd88d59bac5d25b57c0363f
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

cherry-picked This PR was cherry-picked onto a release branch from master open source

Projects

None yet

Development

Successfully merging this pull request may close these issues.

ctc loss get nan after some epochs in pytorch 1.0.0.dev20181115

5 participants