> ## Documentation Index
> Fetch the complete documentation index at: https://docs.nx1cloud.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Create a training job

> Start a training job that fine-tunes a base model on an approved dataset.

You create a training job to fine-tune a base large language model on a dataset you approved. NexusOne starts the
training right away and registers the result in MLflow.

## Prerequisites

Before starting this task, ensure you have the following:

* Required roles: `nx1_finetune_admin`, `nx1_finetune_user`, or `nx1_engineer`.
* A previously created dataset with the `complete` status.
* A base model that your platform team has made available for training.

## Create a training job

> [tab] Web portal
>
> Perform these steps to create a training job:
>
> 1. Log in to NexusOne.
>
> 2. On the top navigation bar, hover your mouse over **AI** and then select **AutoML & AutoLLM**.
>
> 3. Click **Fine-tune LLMs (AutoLLM)**.
>
> 4. Click **Datasets**.
>
> 5. Select the dataset with the `complete` status.
>
> 6. Click **Start training**. NexusOne starts a training job right away, using the QLoRA adapter type.
>
> > [!NOTE]
> >
> > You can also start training automatically when you approve a dataset. To do so, turn on
> > **Start training automatically after approval** before you click **Approve**.

> [tab] REST API
>
> Use the [Create a fine-tune training job](https://docs.nx1cloud.com/api-reference/endpoints/autollm/create-a-fine-tune-training-job.md)
> endpoint to create a training job.
>
> Unlike the web portal, the API lets you choose the adapter type and the number of GPUs
> using the `adapter_type` and `gpu_count` parameters.

## Additional resources

For a high-level overview of training jobs and how they're used in NexusOne, refer to
[Fine-tune LLM](https://docs.nx1cloud.com/documentation/ai/overview/automl/fine-tune-llm.md#training).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.