[Fix][Text Generation Pipeline] Fix the erroneous sampling logic #1406
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
dsikka
reviewed
Nov 15, 2023
|
Could you add output before and after? |
rahul-tuli
approved these changes
Nov 15, 2023
bfineran
approved these changes
Nov 15, 2023
bfineran
pushed a commit
that referenced
this pull request
Nov 16, 2023
bfineran
pushed a commit
that referenced
this pull request
Nov 16, 2023
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Outdated suggestions cannot be applied.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.
Fix Description
Before: regardless of whether
sampling=True or Falsewe would dotop_kandtop_psampling.Now: if
sampling=False, we directly "jump" to the argmax function and avoid any sampling logic.@horheynm Could you please validate the rest of the logic in
def generate(self, logits: numpy.ndarray)? In the most complex scenario, we can apply bothtop_k,top_p, andsampling_temperaturesequentially to our logits. Let's make sure that the order in which the sampling functions are applied matches the one defined in HF (I assume this is the original implementation that we want to mimic).