I’ve run GPT Neo on TPU… It’s stucked in model.generate()
What’s the proper way to run it on TPU?