feat: 2-3x inference speedup, faster than real-time #71
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
vatsalaggarwal
changed the title
sidroopdaska
requested changes
Feb 22, 2024
fam/llm/gptfast_inference.py
Outdated
| num_samples=1, | ||
| seed=1337, | ||
| device="cuda", | ||
| dtype="bfloat16", |
There was a problem hiding this comment.
auto handle dtype. check fam/llm/utils.py
sidroopdaska
approved these changes
Feb 25, 2024
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Outdated suggestions cannot be applied.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.
dirty fix to speed up inference by porting gpt-fast... supports single utterance only, no batching...
First stage only:
RTX 4090: 230T/s (~1.5 seconds of speech generated in 1 second of wall-clock time)
H100: 382T/s (~2.5 seconds of speech generated in 1 second of wall-clock time)
These times are consistent across context lengths due to the static kv-cache used.
Notes: