I’m trying to prepare the input_data for my project but my tensors are of different sizes, for example: [1, 1256], [1, 6497], [1, 58672], etc. I want to have them all of the size [1, 58672]. Can you please tell me how can I do it?
Thanks in advance
Depending how you would like to increase this dimension, you could use
F.pad and pad the dimension to the desired shape
- create another tensor in the “missing” shape and use
torch.cat((x, other), dim=1) to concatenate them
- concatenate the tensor to itself and pad the rest
Let me know, if that would work or if you would like to apply another approach.
Hi ptrblck, thank you for your answer. I’m actually trying to create a Dataset for my audio project but unfortunately I’m still a beginner and I have to learn a lot. I have tried this site: https://pytorch.org/tutorials/beginner/data_loading_tutorial.html but it’s still not working. Can you please suggest me a website where I can learn something more, for example F.pad and torch.cat?
The docs are generally a good starting point.
This code should work for the padding approach:
target_len = 58672
x = torch.randn(1, 1256)
x = F.pad(x, (target_len - x.size(1), 0))
> torch.Size([1, 58672])
Thank you very much, it really helped me