T5X Usage Guides# Pretraining a model Introduction Overview Step 1: Choose a model architecture Step 2: Choose a SeqIO Task/Mixture Step 3: Write a Gin Config Step 4: Launch your experiment Next Steps Advanced Topics Fine Tuning a Model Introduction Overview Step 1: Choose a pre-trained model Step 2: Choose a SeqIO Task/Mixture Step 3: Write a Gin Config Step 4: Launch your experiment Next Steps Advanced Topics Evaluating a Model Introduction Overview Step 1: Choose a model Step 2: Choose a SeqIO Task/Mixture Step 3: Write a Gin Config Step 4: Launch your experiment Step 5: Monitor your experiment and parse metrics Next Steps Advanced Topics Running Inference on a Model Auxiliary Job Introduction Overview Example 1: Separate evaluation job. Example 2: Continuous finetuning job. Common Gotchas. Decoding Option 1: defining a decoding function Option 2: subclassing a model class Metrics Overview Introduction Metrics and Writers Usage in T5X Data, Model, and Activation Partitioning How to Partition Other Stuff Example configurations Gin Primer Gin in T5X Scripts Command-Line Usage Logging
