I was reading your notebook code and noticed that in one line of the code you sample a batch of patients to train the network. Here is the code:
pat_batch = train_pats[torch.randperm(batch_sz)%train_sz][0:batch_sz]
Given from the previous lines that batch_sz=8 and train_sz=63 and train_pats= [0, 1, ..., 62], doesn't it always choose the first 8 patients? Don't you use other 63-8 = 55 patients' volumes?
I was reading your notebook code and noticed that in one line of the code you sample a batch of patients to train the network. Here is the code:
Given from the previous lines that
batch_sz=8andtrain_sz=63andtrain_pats= [0, 1, ..., 62], doesn't it always choose the first 8 patients? Don't you use other63-8 = 55patients' volumes?