Skip to main content

Training Reports

Training available​

Training Reports page shows the list of all trainings performed on the Project.

Training Reports

Details of the training can be seen by selecting a particular training from the list.

ID​

ID denotes an unique identifier which is given to each training.

Algos Trained​

Denotes the number of algorithms trained in the training.

Status​

Status gives the current state of the training, It can be either of

  • Training Done
  • Training in progress
  • Training failed
  • Training aborted
  • Error in fetching training status
  • Evaluation in progress
  • Evaluation Done
  • Error in Evaluation
  • Unknown training status

Success​

Denotes if the current training session has failed or succeeded

Best Trained​

Shows which training has the best algorithm trained with the best score in the project.

Training Details​

When a Training in the training list is selected, the training details is shown, and also the tab to access the training logs.

Training details

Trained algorithms​

This tab shows a table of all the algorithms trained in the training. The table properties are explained

Algorithm​

The name of the algorithm in the training session.

Time taken​

The time taken by the algorithm to train

Status​

Shows the status of the training of the algorithm

Score​

Shows the accuracy of the algorithm in this training

Algorithm Details panel​

Expanding the algorithm will show the details of the algorithm.

Training algorithm details

Algo Details​

Gives details on the selected algorithm

Algorithm details​
  • AlgoID: The unique identifier of the selected algorithm
  • CommitID: The commit hash identifying the latest commit the algorithms repository from where the algorithm is cloned
Algorithm scores​

Shows the Precision, Recall and the F1 score of the algorithm trained

Algo Logs​

Shows the stdout and stderr of the docker container where the algorithm trained.

Reports​

Reports shows individual entity scores and individual record scores of the algorithm.

Entity scores​

Shows the Precision, Recall, F1 score, and Accuracy of the algorithm for individual entities

Training Entity scores

This helps to identify how well the algorithm scores for each entity and which entity needs more attention in the training data to reduce bias.

Per record scores​

Shows the Precision, Recall, F1 score, and Accuracy of the algorithm for individual records Marked for evaluation

Training Per record scores

This helps to identify how well the algorithm scores for each individual records marked for evaluation.

tip

Certain evaluation records might have human errors and sorting the Per record scores on individual parameters and checking the least scoring records can reveal mistakes in human classification which can be rectified to improve the overall score of the algorithm

Artifacts​

Download the artifacts of the trained algorithm. The artifacts of the algorithm are logs and trained model.

Training log​

This tab shows the logs of the training