Hello,
What is the minimum requirements for finetunning and inference? is TPU mandatory? what are the costs/time needed to train a small dataset used in the finetuning article?
This codebase is designed for inference and finetuning on TPU. The cost of TPU instances are available from google, and fine tuning runs at around 5000 tokens per second on a TPU v3-8
Hello, What is the minimum requirements for finetunning and inference? is TPU mandatory? what are the costs/time needed to train a small dataset used in the finetuning article?