Overview of Fine-Tuning Models
The Fine-Tuning Models page provides an interface where users can manage and review the fine-tuned models created. It gives an overview of each model’s status, along with options for further actions and modifications.
Key Features:
- Model Filtering: Users can sort and filter models based on their current status:
All,Successful, orFailed. - Model Overview: Each model displays important information, including the model name, details of the fine-tuned version, and the model’s status.
- Available Actions: Users can interact with models, such as re-running the fine-tuning process or modifying hyperparameters.
Fine-tuning is an essential step in improving a model’s performance by training it with additional data to target specific tasks and enhance accuracy.
How to Get Started:
- Browse Through Existing Models: Browse the list of created models on this page. You can filter by success or failure status, and select any model to perform additional actions.
- Create a New Fine-Tuned Model: To start a new fine-tuning process, simply click on the Create button located in the top-right corner of the page.
Explore the Fine-Tuning Models section to refine model performance according to specific data sets and parameter configurations.
| Field | Description |
|---|---|
| Model Name | The name assigned to the fine-tuned model |
| Fine-Tuned Model | ID and associated details of the fine-tuned model |
| Status | The current status of the model (Succeeded, Failed) |
| Actions | Options for evaluation, deletion, and more |
Fine-Tuning Model Evaluation
The Fine-Tuning Evaluation page gives a detailed breakdown of a fine-tuned model's performance. Users can track checkpoints, review snapshots, and visualize the fine-tuning process.
Evaluation Features:
- Graphical Performance Chart: A visual representation showing the model's performance improvement over the fine-tuning period.
- Training Timeline: A timeline that details when checkpoints were created and when snapshots were saved during the fine-tuning process.
- Detailed Information: Model data such as
Model Name,Created At,Finished At, andFine-Tuned Model IDis displayed for easy reference.
Evaluation data helps identify areas where the model excels and areas that might require more training or adjustments.
Steps to Access Fine-Tuning Evaluation:
- View Evaluation Details: Click the eye icon next to a model on the Fine-Tuning Models page to access detailed evaluation.
- Examine Graph: The evaluation section displays a performance chart showcasing the model’s progress.
- Checkpoints and Snapshots: Detailed information on each checkpoint and snapshot, including step numbers and snapshot identifiers, will be displayed.
| Model Attribute | Details |
|---|---|
| Model Name | The name of the original base model |
| Created At | Timestamp indicating when the fine-tuning started |
| Finished At | Timestamp showing when the fine-tuning was completed |
| Fine-Tuned Model | ID of the fine-tuned model |
| Status | The final status of the fine-tuning (Succeeded or Failed) |
| Error (if any) | Error message or details in case of failure |
Example Page Layout:
On the Fine-Tuning Models page, if no models have been created, a No Data placeholder will be shown. Once models are created, they will be displayed with their respective statuses and available actions.
In the Fine-Tuning Evaluation page, users can track detailed progress information, including checkpoints and performance graphs for better insight into the fine-tuning process.