Training/inference script for tinystories dataset. Once this is done, maybe we'll get to larger training run on multi-gpu, bigger dataset, gradient accumulation, clipping, etc.
Pretrained Generation: Once upon a time it was all about a team of robots or robots and I would have gotten a better joke out of them than this, and when it came to the actual robot, I had no idea what the whole plan was about.
Checkpointed Generation: Once upon a time, there was a girl called Kitty. Grace had an exciting time. One day, Blue was playing in the park. She had a pretty coat and a picture full of a big blue ball. It was an unusual day, and