ml·platform
marin speedrun control panel · Modal A10G

Configure run

Runs a real GPU job on Modal (A10G), evaluates on Paloma, logs to W&B project marin-speedrun. Only one run executes at a time — queued runs start automatically as each finishes.

Live output

idle 
Configure parameters and click "Run now" to start.

Run history

New experiment

base: Marin Speedrun preset

Tick the variable(s) you want to sweep and give each a comma-separated list of values — one run per combination. Everything left unticked stays fixed at its default value.

Tick at least one variable to sweep

Experiments