triton-lang/triton

[Question] Is it possible to cache the autotune output and re-use it later on ?

Open

#1,072 opened on Jan 18, 2023

View on GitHubContributor guide
 (7 comments) (2 reactions) (0 assignees)MLIR (3,126 forks)github user discovery
help wanted

Repository metrics

Stars
 (19,982 stars)
PR merge metrics
 (Avg merge 2d 18h) (185 merged PRs in 30d)

Description

Dear OpenAI Triton Team,

Here is my use case. I am deploying models and the more requests they get, the faster they become. After a large number of requests, it seems the auto-tune converges.

I wonder if that state can be cached and re-used when restarting the server ?

Best, T.C

Contributor guide

Research direction
Investigate whether Triton's autotune output can be serialized and reloaded. Look for any existing caching mechanism in the codebase, and consider implementing a feature to save/load the best configuration. Check related discussions and PRs.
Tech stack
pythonccpp
Domain
backendperformance
Issue type
Research
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
3
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
1-3 hours
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
Active
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
Clear
Prerequisites
PythonCUDA
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
60