fix: Nx.Random.shuffle repeating a single value in certain cases on GPU #1552
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…-nx/nx into pv-fix/random-shuffle-independent
|
I'm still trying to narrow it down, but the issue must be somewhere in the combination of -sort_keys = random_bits(keys[1], shape: tensor.shape)
+sort_keys = randint_split(keys[1], 0, uint32max, shape: tensor.shape)It seems to work just fine. |
|
@polvalente actually, this also fixes it: -sort_keys = random_bits(keys[1], shape: tensor.shape)
+sort_keys = random_bits(keys[1], shape: tensor.shape) |> Nx.as_type(:s32) |
jonatanklosko
approved these changes
Oct 30, 2024
jonatanklosko
changed the title
Co-Authored-By: Jonatan Klosko <[email protected]>
josevalim
pushed a commit
that referenced
this pull request
Nov 16, 2024
…PU (#1552) Co-authored-by: Jonatan Klosko <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
fixes #1551
In short, what was happening is that we were not testing a specific edge case that resulted in an invalid-shaped tensor for the
sort_keyssub-input insort_key_val, that in turn yielded an invalid shuffle when the input is 1D.