Using a trained PytorchWrapper Model

At application time, when you use LF_ApplyClassification or LF_ApplyChunking the following happens:

• If it does not already exist, the LearningFramework copies the python software needed into your data directory into a subdirectory with the name FileJsonPytorch
• the PR starts the application program in the software directory and establishes a connection (through a pipe)
• the PR processes all the documents in the corpus and passes all the instances in the documents to the application program.
• The application program passes the prediction back to the PR which then uses the PR to set the feature or prepare the chunking