The trainer runs various resource-intensive model-training jobs that, if run on the same machine as the application, would slow down submission processing. The trainer runs separately from the main application but connects to it via the API.
Processing Semi-structured documents, such as invoices, paystubs, and explanation of benefits, requires installing the Hyperscience Trainer on a dedicated machine with at least 64GB of RAM. We recommended that all customers install the trainer to take advantage of Transcription Accuracy Training and Potential Layout discovery.
See the following links for more information on all of the jobs that run in the trainer:
Navigating the Trainer Tab
Navigate to Administration > Trainer to view and manage all running jobs.
On the Trainer page, you can:
View all running, pending, failed, finished, and canceled jobs.
Cancel pending and running jobs.
For more information on this, see Canceling or Retrying a Training Job.
View how many and which trainers are connected to the application.
If no trainer model is detected, a warning message will show at the top of the Trainer Queue.
If an incompatible trainer model is connected (e.g., a trainer version that is older than the version of the application), the UI will indicate this incompatibility. You should uninstall the incompatible trainer because it will use up memory even while idle.
Guidelines for Running Trainer
Separate from the main application, you must also install the trainer application on a dedicated machine with at least 64GB of RAM. You can also run multiple trainers on one machine, but only one job will run at a time per machine.
You can have multiple trainers on multiple machines to run different jobs simultaneously, including two trainers of the same version. Only one job can run at a time per machine.
All jobs are cancellable in the Trainer Queue.
If a job fails, the trainer will retry it two additional times before entering a terminal failed state.