Skip to content

Conversation

TroyGarden
Copy link
Contributor

Summary:

context

  • the original implementation of "ir_custom_op" strategy has logic flaw:
  • input the sum of dim, and let the op return a contiguous tensor, then split it to multiple tensors
  • from the dynamic shape (ds) prespective, there is a sum(ds_i) before the op, then another split to (ds_i). the range calculation for these ds are unnecessary and create a lot of complexities
  • it's better to keep these ds transparent into and out from the op

Differential Revision: D53558783

@facebook-github-bot facebook-github-bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Jul 26, 2024
@facebook-github-bot
Copy link
Contributor

This pull request was exported from Phabricator. Differential Revision: D53558783

Summary:
Pull Request resolved: pytorch#2246

# context
* the original implementation of "ir_custom_op" strategy has logic flaw:
* input the sum of dim, and let the op return a contiguous tensor, then split it to multiple tensors
* from the dynamic shape (ds) prespective, there is a sum(ds_i) before the op, then another split to (ds_i). the range calculation for these ds are unnecessary and create a lot of complexities
* it's better to keep these ds transparent into and out from the op

Differential Revision: D53558783
@facebook-github-bot
Copy link
Contributor

This pull request was exported from Phabricator. Differential Revision: D53558783

@TroyGarden TroyGarden deleted the export-D53558783 branch August 8, 2024 22:14
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. fb-exported
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants