API reference · Fine-tuning
A job's training metrics, one point per step in step order (chart-ready)
/v1/fine-tunes/{fine_tune_id}/metrics- TypeScript
client.fineTuning.listMetrics()- Python
client.fine_tuning.list_metrics()- CLI
hayonix fine-tunes metrics
Up to 200 points per page by default. While the job trains, poll with after_step set to the last step you have to get only new points.
Project API key (Authorization: Bearer sk_…), or a console session. Requires the fine_tune_job:read permission (a key's scope, or a role that includes it).
Examples
The samples read API_KEY and API_BASE_URL from the environment (see Setup). Values are examples.
curl -sS "$API_BASE_URL/fine-tunes/ftj_01jc5zq4m8k3v6x9t2w7r0b1e4/metrics" \
-H "Authorization: Bearer $API_KEY"Request
| Header | Required | Description |
|---|---|---|
Authorization | yes | Bearer and your API key. The key implies its project. |
X-Project-Id | session only | The project to act in, for console sessions. Requests with an API key leave it out. |
Path parameters
fine_tune_idstringrequiredID of the fine-tuning job
ID with prefix ftj_
Query parameters
limitinteger1–200, default 50
1–200
cursorstring`next_cursor` of the previous page
up to 200 characters
after_stepinteger | nullOnly points after this step (polling)
at least -1
Response
200 a page of results — A page of metric points
The items of this page.
ShowHide 9 fields of
dataobjectstringrequiredAlways `fine_tune_metric_point`: the type of this object
One of
fine_tune_metric_pointstepintegerrequiredOptimizer step the values were recorded at
epochnumber | nullrequiredEpoch at this step, fractional; null when not reported
train_lossnumber | nullrequiredTraining loss; null when not reported at this step
eval_lossnumber | nullrequiredLoss on the validation file; null at steps without an evaluation
learning_ratenumber | nullrequiredLearning rate at this step; null when not reported
grad_normnumber | nullrequiredGradient norm before clipping; null when not reported
valuesobjectrequiredEvery metric at this step (preference training adds reward metrics)
recorded_atstring (date-time)requiredTime recorded (ISO 8601, UTC)
next_cursorstring | nullrequiredPass as `cursor` for the next page; null on the last page.
Example response
{
"data": [
{
"object": "fine_tune_metric_point",
"step": 1,
"epoch": 1.5,
"train_loss": 1.5,
"eval_loss": 1.5,
"learning_rate": 1.5,
"grad_norm": 1.5,
"values": {},
"recorded_at": "2026-01-01T00:00:00.000Z"
}
],
"next_cursor": "string"
}Errors
Errors use the standard envelope. What this operation can answer:
| Status | Type | When |
|---|---|---|
400 | invalid_request_error | A parameter or body field is missing or not allowed; param names it. |
401 | authentication_error | The key or session is missing, revoked or expired. |
403 | permission_error | The credential lacks the permission this needs, or the organization is suspended. |
404 | not_found_error | No such resource in this project (an ID from another project is answered the same way). |
429 | rate_limit_error rate_limited | Too many requests; retry after retry-after seconds. |