Open zgchen33 opened 5 months ago
Did you use FP16 to train your model? And was the performance in the paper trained by FP32?
All the models were trained by FP32
Thank you for your response. By the way, I don't see any checkpoint you released. Do you plan to release the checkpoints?
Did you use FP16 to train your model? And was the performance in the paper trained by FP32?