Closed BingY998 closed 7 months ago
In this code, the main role of the number of GPUs is to accelerate the training. only one GTX3090 with 24G memory is okay to train LAfomer, but maybe you need to reduce the batch size to one-half or one-quarter of the current and then increase the learning rate accordingly.
Thank you very much. I will do it!
Hello, thank you very much for your excellent work! I would like to ask if it is possible to reproduce this code if there is only one GTX3090 with 24G memory?
My computer is currently configured the same as yours, is it convenient to communicate with you? Can you leave a contact information on your homepage?
Hello, thank you very much for your excellent work! I would like to ask if it is possible to reproduce this code if there is only one GTX3090 with 24G memory?