Fine-tune Service¶

Overview¶

The Fine-tune Service on TurboAI Cloud provides a platform for fine-tuning pre-trained models with custom datasets. This service is designed to help users adapt models to specific tasks or domains, enhancing their performance and accuracy.

Key Features¶

• Custom Dataset Integration: Easily upload and integrate your datasets for fine-tuning.
• Parameter Customization: Flexibly set fine-tuning parameters such as learning rates and epochs.
• Performance Evaluation: Utilize tools to evaluate the performance of fine-tuned models to ensure they meet your standards.
• Resource Optimization: Leverage TurboAI Cloud's infrastructure to optimize resource allocation for your fine-tuning tasks.

Accessing Fine-tune Service¶

To access the Fine-tune Service, navigate to the TurboAI platform and select "Fine-tune" from the Training Service menu.

Creating a New Fine-tune Job¶

Step 1: Name¶

Please enter the name of your fine-tune job.

Step 2: Model¶

Select a pre-trained model from the dropdown menu.

Step 3: Method¶

Choose the fine-tuning method, either "Full" or "Lora".

Step 4: Epoch¶

Set the number of training epochs.

Step 5: Learning Rate¶

Set the learning rate, for example, 0.00001.

Step 6: Data Path¶

Upload the data path that you have stored in object storage.

Step 7: Validation Ratio¶

Set the proportion of the dataset to be used for validation, for example, 0.01.

Step 8: Resources¶

Select the location of the resources you wish to use for the fine-tune job, such as Malaysia.

Step 9: Confirm¶

Review the estimated cost and confirm the creation of the fine-tune job.

Fine-tune Job List¶

In the fine-tune job list, you can view the name, method, model, status, resource, creation time, and operation options for all fine-tune jobs.

Next Steps¶

After completing the fine-tuning, you can proceed to explore the Training Service to continue your model development process.