Sampling optimizations #152
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
Code Metrics Report─────────────────────────────────────────────────────────────────────────────── Language Files Lines Blanks Comments Code Complexity ─────────────────────────────────────────────────────────────────────────────── Rust 60 20036 1439 820 17777 1130 ─────────────────────────────────────────────────────────────────────────────── Total 60 20036 1439 820 17777 1130 ─────────────────────────────────────────────────────────────────────────────── Estimated Cost to Develop 54,557 Estimated Schedule Effort 10.992939 months Estimated People Required 4.481762 ─────────────────────────────────────────────────────────────────────────────── Processed 677983 bytes, 0.678 megabytes (SI) ─────────────────────────────────────────────────────────────────────────────── |
lucasavila00
commented
Apr 16, 2024
| // Sort by descending probability. | ||
| argsort_indices.sort_by(|&i, &j| probs[j].partial_cmp(&probs[i]).unwrap()); | ||
| argsort_indices.sort_unstable_by(|&i, &j| probs[j].partial_cmp(&probs[i]).unwrap()); |
There was a problem hiding this comment.
This sort is unstable (i.e., may reorder equal elements), in-place (i.e., does not allocate), and O(n * log(n)) worst-case.
The benchmarks here show the difference between stable and unstable. (I don't think adding the lib is required)
https://github.com/orlp/glidesort
This saves 500ms per token
lucasavila00
commented
Apr 16, 2024
| for (token_id, logit) in logits.iter_mut().enumerate() { | ||
| let count = context.iter().filter(|x| **x as usize == token_id).count(); | ||
| *logit = *logit | ||
| - count as f32 * frequency_penalty | ||
| - if count > 0 { 1. } else { 0. } * presence_penalty; | ||
| } | ||
| let logits_len = logits.len(); | ||
| Tensor::from_vec(logits, logits_len, device) | ||
| Ok(logits) |
There was a problem hiding this comment.
Since this reads the context we can't pre-allocate the tensor.
Allocating the tensor here was triggering a synchronization that would cost 500ms
EricLBuehler
approved these changes
Apr 16, 2024
|
Thank you! |
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.
Makes sampling fully async, removing synchronizations.
Generation goes from 54t/s to 59t/s using
./target/profiling/mistralrs-server --prompt "Tell me 3 jokes." mistral-ggufSame generation speed as Llama.cpp 😄