-
Notifications
You must be signed in to change notification settings - Fork 892
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Implement top_k sampling #13
Merged
Merged
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
This reverts commit e4b7b98.
This allows infer() to run with significantly less vram.
|
Looks good, can you just confirm that it doesn't break if you set top_k=None in infer? |
|
Yes I tested that and it works fine 👍 |
|
Thanks! |
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Outdated suggestions cannot be applied.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.
This seems to work quite well in combination with top_p, I posted some comparisons on discord.
Using "hk.with_rng" to keep the seed consistent across runs makes it much easier to compare changes in the sampling methods, I used this to compare with and without top_k:
Prompt
top_p=0.9, seed=0
top_p=0.9+top_k=50, seed=0
top_p=0.9, seed=1
top_p=0.9+top_k=50, seed=1
It seems to stay on topic better with different prompts so I've enabled it by default.
top_k can be set to None to disable. I wanted to implement this using slices, but couldn't get it working properly. jax.lax.dynamic_slice isn't playing nicely. However, jnp.where produces identical results