Training Services¶

Overview¶

Training Services on TurboAI Cloud are designed to provide a comprehensive platform for training and fine-tuning machine learning models. This includes a range of tools and infrastructure that enable users to efficiently train models from scratch or adapt pre-trained models to specific tasks or domains.

Key Components¶

Training Service¶

• Description: A robust platform for training models from scratch using your own custom datasets.
• Features:
• User-friendly interface for managing machine learning model training tasks.
• Deploy, monitor, and manage training jobs with ease.
• Provision GPU resources specifically designed for high-efficiency training.

Finetune Service¶

• Description: A simple and efficient way to customize pre-trained models with your own data.
• Features:
• Ideal for transfer learning.
• Leverage knowledge from pre-trained models and adapt to specific use cases.
• Significantly reduce training times and improve accuracy for niche applications.

Getting Started¶

Training Service¶

1. Log in to the Console: Navigate to the Training section.
2. Create a New Task: Click the "New" button to create a new training task.
3. Configure Parameters: Set model, workers, shared memory, and resources.
4. Deploy Task: Review and deploy your training task to begin the training process.

Finetune Service¶

1. Name Your Job: Enter the name of your fine-tune job.
2. Select Model: Choose a pre-trained model from the dropdown menu.
3. Set Parameters: Choose the fine-tuning method, set epochs, and define the learning rate.
4. Upload Data: Upload the data path for your custom dataset and set the validation ratio.
5. Select Resources: Choose the resources for your fine-tune job and confirm the creation of the job.

Next Steps¶

After deploying your training tasks or fine-tune jobs, you can:

• Monitor their progress and performance using the respective dashboards.
• Adjust configurations as needed to optimize results.
• For further assistance or to explore advanced features, refer to the related sections or visit our support page.