If this issue occurs, the dataset version is successfully released but does not meet the requirements of the ExeML training jobs. As a result, an error message is displayed, indicating that the dataset version does not meet the requirements.
Labeling Information Does Not Meet the Trainning Requirements
For different types of ExeML projects, training jobs have the following requirements on datasets:
- Image classification: There are at least two classes (that is, at least two labels) for the images to be trained, and the number of images in each class cannot be less than 5.
- Object detection: There is at least one class (that is, at least one label) for the images to be trained, and the number of images for each class cannot be less than 5.
- Predictive analytics: The dataset of the predictive analytics task is not managed in a unified manner. Even if the data does not meet the requirements, no fault information is displayed in this issue.
- Sound classification: There are at least two classes (that is, at least two labels) for the audio files to be trained, and the number of audio files in each class cannot be less than 5.
- Text classification: There are at least two classes (that is, at least two labels) for the text files to be trained, and the number of text files in each class cannot be less than 20.