I found that if I restart the kernel so I can train much faster. I think the reason is because of the GPU memory. GPU memory not being freed after training is over . I will read to see how to free the memory without restarting the kernel
1 Like