RazorThink Platform Basics - Create a Deep Learning Model
Modified on: 2021-07-29 14:06:53 +0530
As an RazorThink Platform User or Admin, you use the RazorThink DL Model Designer to create a Deep Learning (DL) model.
Roles for performing this task
- Log in to RazorThink Platform as a User or Admin. Open the project that will contain the DL model you want to create.
- In the toolbar on the left side of the window, click the DL Model Designer icon . Then DL model list is displayed.
- Click the Plus iconto create a new DL model. The Create Model dialog is displayed.
- In the Create Model dialog Name field, specify a name for the DL model (alphanumeric characters and the underscore may be used; other special characters including space are not valid). Click Create.
The new model is created and shown in the Model List.
- Click the model name in the Model List to begin building the model architecture. Click the Plus icon to add layers, costs, optimisers, and other elements.
- In the Search field, start typing the name of the layer, cost, optimiser, or operator you want to insert in the Designer canvas. A list of available items is displayed.
- Click the layer name and place the layer in the Designer main window. Edit the name of the component.
- Click the component and set parameters of the component.
- Repeat this process for all required components. Connect the output of each layer to the input of the next layer by placing the mouse pointer on the respective output/input dents and dragging from output to input dent. To delete a connection or layer, right-click on the connection/layer and click the Delete option.
- Save the model by clicking Save in the top right corner. The system validates the name and properties of all layers. Error messages are shown on the top right panel if:
a) the tensor shapes are not matching
b) any of the input/output are left unconnected, or
c) any required components such as optimiser are missing.
- Click the Error icon to see all error messages.
- Fix errors and save the final model.
Can you please tell us how we can improve this article?