MLCommons · AI · Instrument: Prize
AlgoPerf: Training Algorithms Benchmark Competition
MLCommons ran the AlgoPerf: Training Algorithms benchmark competition from 28 November 2023 to 28 March 2024 to measure neural network training speedups from algorithmic improvements such as better optimizers and hyperparameter tuning protocols. MLCommons funded a $50,000 prize pool for the top submissions in each of two tuning rulesets, external tuning and self-tuning, and Google provided compute to score submissions. The August 2024 results report 18 submissions from 10 teams. The $25,000 external tuning prize went to the Distributed Shampoo submission by researchers at Meta. The self-tuning prize was not awarded because the affiliation conditions were not met.
Who can apply
- Eligibility
No separate eligibility field is recorded. Check the program summary and the funder's terms.
- Location
No separate location field is recorded. Check the program summary and the funder's terms.
- Funding
Funding classification
- Instrument
No separate instrument field is recorded. Check the program summary and the funder's terms.
- Capital form
No separate capital form field is recorded. Check the program summary and the funder's terms.
- Allocation method
No separate allocation method field is recorded. Check the program summary and the funder's terms.
- Selection
No separate selection field is recorded. Check the program summary and the funder's terms.
Round history
| Round | Window | Size | Source |
|---|---|---|---|
| Inaugural competition | November 28, 2023 to March 28, 2024Closed | $50,000 prize pool; $25,000 awarded | mlcommons.org |
Past rounds stay on the page, so you can see how often this program opens.
Funding evidence
Detailed funding records are not yet recorded for this program.
Read the program’s recorded facts and sourcesSources
Awards
none linkedNo award records linked to this profile yet.