Introduction to Automatic Modeling

Here we introduce the use of HyperAI Automatic modeling module. Developers can quickly train a deep learning model. Solving corresponding deep learning problems. Developers need to convert the format of their dataset into HyperAI data format. Then use the automatic modeling module according to the steps. By using the corresponding default model or automatic parameter tuning based on the default model. Obtain the model with the best metric.

Quick try

For the convenience of explanation. We will use the object detection problem as an example to illustrate its usage. And use a publicly available dataset for automatic modeling functionality. Select the left side“Automatic modeling”, And click“Create a new automatic modeling”Button. Start an automatic modeling job.

Create a new automatic modeling

Enter the new automatic modeling interface. Fill in corresponding information. And in“solve the problem”In one item, choice“object detection ”problem.

Bind data

We have prepared a public dataset“defect detection ”Used to showcase. In binding data projects, input0 It is a publicly available pre trained model prepared for automatic modeling. Can accelerate the training process; input1 It is the default selected defect dataset. During use. Can be adjusted according to one's own needs input1 The binding object.

To configure

On the configuration interface. The information from the previous page will be displayed above. This section is only for display purposes. Cannot be modified. The lower part is the configuration options, “Choose computing power”What is currently provided is NVIDIA V100 Graphics card and NVIDIA A100 The computing power type of the graphics card; “Maximum attempts”It refers to the process of automatic modeling. The total number of attempts during the learning process. Automatic modeling will find the optimal model within this range within a specified number of times; “Parallel number”It refers to simultaneous activation GPU Number of training instances. This parameter is affected by personal permissions.“advanced setting”The initialization parameters in are set parameters related to the dataset. The user must fill in the width and height of the dataset images. After completing the configuration, click“function”, Start automatic modeling job.

Result Display

After starting to run. Until all execution attempts are completed. We can view the basic information of the task through the overview interface. Execution status and results of each attempted execution.