There are multiple problems that we are seeing with the training
First, the MultVAE is slower than MultiBatchVAE, which makes no sense, since MultiBatchVAE has many more parameters and is more complex.
Second, we are noticing that TripletNets randomly get hung up in the middle of training. There is likely something messed up going on with the dataloader workers.
There are multiple problems that we are seeing with the training
First, the MultVAE is slower than MultiBatchVAE, which makes no sense, since MultiBatchVAE has many more parameters and is more complex.
Second, we are noticing that TripletNets randomly get hung up in the middle of training. There is likely something messed up going on with the dataloader workers.