Axom Docs

Quickstart ​

Go from zero to a running Fusion job. You'll create a key, point at a dataset and a base model, submit a run, and watch its telemetry.

1. Create an API key ​

In the Console → API keys, create a key and copy the secret (shown once). See Authentication for details.

bash
export AXOM_KEY="axom_live_…"

2. Submit a training run ​

bash
curl https://console.axomlabs.ai/api/jobs \
  -H "Authorization: Bearer $AXOM_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "type": "train",
    "idempotency_key": "my-first-run-1",
    "dataset_id": "<dataset-id>",
    "hyperparams": { "base": "Llama-3-8B", "cycles": 6, "steps_per_cycle": 1000 }
  }'

The response includes the new job's id and status (queued).

3. Watch it train ​

Poll the per-step metrics, fetching only points newer than the last you have:

bash
curl "https://console.axomlabs.ai/api/jobs/<job-id>/metrics?since_step=0" \
  -H "Authorization: Bearer $AXOM_KEY"
json
{ "job_id": "…", "count": 50, "latest_step": 1000,
  "points": [ { "step": 20, "loss": 2.94, "lr": 5.5e-05, "vram_gb": 23.3 }, … ] }

4. Get the result & download the model ​

When the run completes, fetch the result summary, then download the trained model. The download endpoint returns a presigned URL per file — you pull the weights directly from storage.

bash
# headline result: perplexity, params, cost
curl https://console.axomlabs.ai/api/jobs/<job-id>/result -H "Authorization: Bearer $AXOM_KEY"

# your models, then a download manifest for one
curl https://console.axomlabs.ai/api/models -H "Authorization: Bearer $AXOM_KEY"
curl https://console.axomlabs.ai/api/models/<model-id>/download -H "Authorization: Bearer $AXOM_KEY"

Next: the full API Reference · Pricing.

Fusion Training Console

© Axom Labs