Full Pipelines

A Full Pipeline runs a Training Pipeline and an Evaluation Pipeline together.

Important:

Minimal dataset size

For successfully running a Training pipeline, we strongly recommend at least 25 documents and at least 10 samples from each labeled field in your dataset. Otherwise, the pipeline throws the following error: Dataset Creation Failed.

Training on GPU vs CPU

• For larger datasets, you need to train using GPU. However, using a GPU (AI Robot Pro) for training is at least 5 times slower than using a CPU (AI Robot).
• Training on CPU is only supported for datasets up to 5000 pages in size for ML Packages v21.10.x and up to 1000 pages for other versions of ML Packages. Training on CPU instead of GPU could also lead to a less accurate model (between 0-5%).
• CPU training was limited to 500 pages before 2021.10, it went up to 5000 pages for 2021.10 and with 2022.4 it will come back down to 1000 pages max.

Train and Evaluate a Model at the Same Time

Configure the training pipeline as follows:

• In the Pipeline type field, select Full Pipeline run.
• In the Choose package field, select the package you want to train and evaluate.
• In the Choose package major version field, select a major version for your package.
• In the Choose package minor version field, select a minor version for your package. It is strongly recommended to always use minor version 0 (zero).
• In the Choose input dataset field, select a representative training dataset.
• In the Choose evaluation dataset field, select a representative evaluation dataset.
• In the Enter parameters section, enter any environment variables defined, and used by your pipeline, if any. For most use cases, no parameter needs to be specified; the model is using advanced techniques to find a performant configuration. However, here are some environment variables you could use:
• auto_retraining which allows you to complete the Auto-retraining Loop; if the variable is set to True, then the input dataset needs to be the export folder associated with the labeling session where the data is tagged; if the variable remains set to False, then the input dataset needs to correspond to the following dataset format.
• model.epochs which customizes the number of epochs for the Training Pipeline (the default value is 100).
• Select whether to train the pipeline on GPU or on CPU. The Enable GPU slider is disabled by default, in which case the pipeline is trained on CPU.
• Select one of the options when the pipeline should run: Run now,Time based or Recurring. In case you are using the auto_retraining variable, select Recurring.
  
  

  

  
  

After you configure all the fields, click Create. The pipeline is created.

Artifacts

For an Evaluation Pipeline, the Outputs pane also includes an artifacts / eval_metrics folder which contains two files:

• evaluation_default.xlsx is an Excel spreadsheet with a side-by-side comparison of ground truth versus predicted value for each field predicted by the model, as well as a per-document accuracy metric, in order of increasing accuracy. Hence, the most inaccurate documents are presented at the top to facilitate diagnosis and troubleshooting.
• evaluation_metrics_default.txt contains the F1 scores of the fields which were predicted.

  For line items, a global score is obtained for all columns taken together.