InstantMLDocs

InstantML Docs

Log, compare, and debug training runs with the InstantML SDK, API, and dashboard.

Open .md

InstantML is hosted training observability for small ML teams: log runs from your Python training loop, then sort, chart, and compare them in the dashboard. Start with the quickstart to go from API key to a charted run in about five minutes, or jump straight to the area you need below.

QuickstartGo from API key to your first charted run in five minutes.Install the SDKInstall the Python package and authenticate.Log from PythonCreate runs, log metrics, and attach configs and tags.Compare runsChart runs side by side and sort by metric summaries.Dashboard tourWalk through the main product surfaces with screenshots.WorkspacesManage members, roles, billing scope, and workspace switching.ReportsWrite experiment reports with live panels and share links.API referenceBrowse the public REST API reference.ExamplesRun the bundled supervised, RL, checkpoint, and distributed examples.Usage and limitsUnderstand exports, usage counters, and plan limits.PricingCompare Free, Pro, and Premium plans and what each includes.

Log your first run

Log a complete run with a project, config, tags, and three metric series:

python
import instantml as im

run = im.init(
    project="cartpole",
    name="ppo-seed-42",
    config={"seed": 42, "algorithm": "ppo", "learning_rate": 3e-4},
    tags=["ppo", "baseline"],
    notes="First PPO baseline.",
)

for step in range(100):
    run.log(
        {
            "train/reward": step * 1.5,
            "train/loss": 1.0 / (step + 1),
            "eval/accuracy": min(1.0, 0.5 + step / 200),
        },
        step=step,
    )

run.finish()

Open the dashboard when the script finishes. The project appears in the project selector, the run appears in the run list, config values appear in Run Detail and Compare, and every metric key becomes a chartable series.

The Runs workspace charting logged metric series across runs
The Runs workspace charting logged metric series across runs

Ready to run it yourself? The quickstart walks through the API key, install, and first run end to end, and Core concepts explains the workspace, project, run, and metric model behind it.