Skip to main content
You approve a dataset to confirm that its example conversations are ready for fine-tuning. Nothing trains on a dataset until you approve it.

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 pending_approval status.

Approve a dataset

Perform these steps to approve a dataset:
  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 pending_approval status.
  6. Review the sample of example conversations. Each shows a question and an answer.
    The dialog shows only a sample of the examples, so only the examples in that sample have checkboxes. To read every example before you approve, refer to View a dataset.
  7. Optional: Select the checkbox next to each example you want to leave out. NexusOne doesn’t remove them yet. It drops the selected examples when you click Approve in step 9, and the button shows how many.
  8. Optional: Turn on Start training automatically after approval to have NexusOne start training as soon as it finalizes the dataset. The automatic option always uses the QLoRA adapter type.
  9. Click Approve.
After you approve the dataset, its status changes from approved to materializing and then to complete. When it’s complete, NexusOne has saved the final dataset and it’s ready to train on.

Additional resources

For a high-level overview of datasets and how they’re used in NexusOne, refer to Fine-tune LLM.