Francesco
Francesco
Could you please show the script you're using? Are you trying to use the pretrained models? The issue you're seeing is due to the system trying to use GPU, but...
Hi, the whitespace collapse is a wanted effect, mostly to be able to control where the pauses are allocated with the forward model. You can remove this if you want...
I used to, but I could not notice a difference in the results so to simplify the structure and to make it closer to the pre-trained vocoders I removed it.
I think I use a p2.xlarge for training (about a day of training I believe). The model is automatically saved in the weights directory.
The list of images seems empty, are you pointing to the right directory?
Hard to say, how many data points are you using for training and test? Is PSNR increasing for training set? What config are you using?
Yes, it does make a lot of sense, it is kind of the purpose of this project. There is no automatic way to do cross experiment selection, but set up...
When doing inference you should not specify an input size when creating a model instance. For example `model = RRDN(arch_params={'C':4, 'D':3, 'G':32, 'G0':32, 'T':10, 'x':2})` If this is not your...
> You're right, removed the patch size and model predicts successfully. > Another issue arise now when i train the model, i receive similar error which comes at random phase...
> Yes, it does, printing out lr_img shows a numpy array. > I have now tried the exact same thing with a much lower resolution jpg, instead of a bigger...