> ## Documentation Index
> Fetch the complete documentation index at: https://docs.valkyrieapp.azumo.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Fine-Tuning screen

> Create and monitor training jobs from the dashboard.

The **Fine-Tuning** screen is where you create training jobs, watch their progress,
and get to their outputs. For the concepts behind it, see the
[Fine-tune a model](/guides/fine-tuning) guide.

<Frame caption="The Fine-Tuning Jobs list.">
  <img src="https://mintcdn.com/valkyrie-azumo/gcj5Fb5HjPMoSXfu/images/dashboard/fine-tuning.png?fit=max&auto=format&n=gcj5Fb5HjPMoSXfu&q=85&s=1dc0aed11431403650bc998332a49959" alt="Valkyrie Fine-Tuning Jobs screen" width="2880" height="1800" data-path="images/dashboard/fine-tuning.png" />
</Frame>

Each job card shows its base model, dataset, key training settings (epochs, batch
size), status, and a **View job** link for logs and outputs.

## Create a job

<Steps>
  <Step title="Start a new job">
    Choose the open base model to fine-tune and the [dataset](/guides/datasets) to
    train on.
  </Step>

  <Step title="Launch">
    Valkyrie schedules the job, provisions a GPU, and runs the training.
  </Step>
</Steps>

## Monitor

Jobs show their current state as they progress:

```
pending → scheduled → running → completed / failed
```

You can stream logs while a job runs to follow training progress.

## Outputs and next steps

When a job completes, open it to:

* **Download outputs** (trained weights or adapters) via temporary signed URLs.
* **Deploy** the fine-tuned model directly to a serving endpoint.

## Recover a failed job

Because jobs run on ephemeral cloud GPUs, transient failures happen. Use **retry** or
**recover** to reschedule a job onto fresh hardware without recreating it.