> ## 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.

# Datasets

> Upload and manage the datasets you fine-tune on.

Datasets are the training data for your fine-tuning jobs. You upload a dataset once
and reference it from any number of jobs.

<Frame caption="The Datasets screen: name, files, size, and dataset ID.">
  <img src="https://mintcdn.com/valkyrie-azumo/gcj5Fb5HjPMoSXfu/images/dashboard/datasets.png?fit=max&auto=format&n=gcj5Fb5HjPMoSXfu&q=85&s=f9b65cdbe3cb600c3881c1f104daa27f" alt="Valkyrie Datasets screen" width="2880" height="1800" data-path="images/dashboard/datasets.png" />
</Frame>

## Upload a dataset

<Steps>
  <Step title="Open Datasets">
    In the dashboard, go to **Datasets** and choose to add a new dataset.
  </Step>

  <Step title="Upload your file">
    Provide your dataset file. Valkyrie issues a secure upload URL and stores the
    file for you.
  </Step>

  <Step title="Reference it in a job">
    When you create a [fine-tuning job](/guides/fine-tuning), select this dataset as
    the training data.
  </Step>
</Steps>

## Formatting your data

Fine-tuning expects structured conversational or instruction data. A common,
portable shape is JSON Lines (`.jsonl`), one training example per line, for example
a chat-style record:

```json theme={null}
{"messages": [{"role": "user", "content": "..."}, {"role": "assistant", "content": "..."}]}
```

<Tip>
  Keep examples consistent in structure and representative of how the model will be
  prompted in production. Quality and consistency matter more than raw volume.
</Tip>

## Managing datasets

From the Datasets screen (or via the [MCP tools](/mcp/tools)) you can:

* **List** your datasets and inspect their metadata.
* **Download** a dataset with a temporary signed URL.
* **Delete** a dataset you no longer need.

<Note>
  Datasets are stored in object storage. Deleting a dataset does not affect models
  you've already trained from it.
</Note>