I have a list and there are many tensor in the list I want to turn it to just tensor, and I can put it to dataloader I use for loop and cat the tensor but it is very slow, data size is about 4,800,000

check this out but summary use torch.stack if you want to respect the original nesting of the lists (by having a tensor with many indices and dimensions) and incrementally create it. There might be better ways but that works for me. [image] Best way to convert a list to a tensor? …