Should I use pytorch or albumentations totensor functions in dataset?


I’ve noticed there is a shape mismatch when I use albumentations to tensor (channel gets appended in the mask).
I.e. I get batch_size, height, width, channels
When I use pytorch transforms to tensor I get the correct batch_size, channels, height, width for both images and masks.
Now I’m transposing, but I’m wondering if this is the wrong approach?