Skip to main content
GET
/
fine-tune
/
jobs
/
{job_id}
/
metrics
curl "https://api.u1.archetypeai.app/v0.5/fine-tune/jobs/ft-job-my-finetune-run-3de60a5c/metrics?limit=100" \
  -H "Authorization: Bearer $ATAI_API_KEY"

{
  "last_checkpoint_step": 600,
  "metrics": [
    {
      "step": 200,
      "loss": 1.842,
      "eval_loss": 1.901,
      "learning_rate": 0.0002,
      "wall_time_sec": 312.4
    },
    {
      "step": 400,
      "loss": 1.413,
      "eval_loss": 1.502,
      "learning_rate": 0.00018,
      "wall_time_sec": 622.7
    },
    {
      "step": 600,
      "loss": 1.187,
      "eval_loss": 1.298,
      "learning_rate": 0.00015,
      "wall_time_sec": 935.1
    }
  ]
}

​
Overview

This endpoint returns the per-step training and evaluation metrics emitted by a fine-tune job. Metrics are recorded at each checkpoint step; use the limit and offset query parameters to paginate through them.

​
Path Parameters

​
job_id
string
required
Identifier of the job whose metrics to fetch (1–64 characters)

​
Query Parameters

​
limit
integer
default:"-1"
Maximum number of metric entries to return. Use -1 (the default) to return all entries.
​
offset
integer
default:"0"
Number of metric entries to skip before returning results. Use with limit to paginate.

​
Response

​
last_checkpoint_step
integer
The most recent checkpoint step recorded for the job. -1 when no checkpoint has been written yet. Useful for tracking progress without paginating through all metrics.
​
metrics
array
Array of metric records. Each record is a free-form object emitted by the runner — typical keys include step, loss, eval_loss, learning_rate, and timing information.
curl "https://api.u1.archetypeai.app/v0.5/fine-tune/jobs/ft-job-my-finetune-run-3de60a5c/metrics?limit=100" \
  -H "Authorization: Bearer $ATAI_API_KEY"

{
  "last_checkpoint_step": 600,
  "metrics": [
    {
      "step": 200,
      "loss": 1.842,
      "eval_loss": 1.901,
      "learning_rate": 0.0002,
      "wall_time_sec": 312.4
    },
    {
      "step": 400,
      "loss": 1.413,
      "eval_loss": 1.502,
      "learning_rate": 0.00018,
      "wall_time_sec": 622.7
    },
    {
      "step": 600,
      "loss": 1.187,
      "eval_loss": 1.298,
      "learning_rate": 0.00015,
      "wall_time_sec": 935.1
    }
  ]
}