Is it good practice to return a random sample in getitem?
My use case is:
I am doing training over videos, and I want to return random subsequences of videos within each minibatch.
I’m not sure the best way to do this.
i have not trained a model on videos. But I trained models on images. I always set shuffle True and set random torch.manual_seed(10). So that the random sequence is the same for each time.