Search⌘ K
AI Features

Training Sampler

Explore the decoding process during Seq2Seq model training by learning to create a TrainingSampler object. Understand how to use ground truth embeddings and sequence lengths to guide decoding, preparing you to implement effective training decoders in NLP tasks.

Chapter Goals:

  • Learn about the decoding process during training

  • Create a TrainingSampler object to use for decoding

A. Decoding during training

During training, we have access to both the input and output sequences of a training pair. This means that we can use the output sequence's ground truth tokens as input for the decoder. ...