Dat Duong

I have the same question here about taking softmax to compute attention weights. I rewrote my code to explicitly truncate each sample in the batch (quite inefficient). Some preliminary result...

This line here still does not use any masking https://github.com/jamesmullenbach/caml-mimic/blob/master/learn/models.py#L184 to compute weights.

It looks like the input tensor (the image) has to be in shape channel x width x height. However, this code may take only grayscale images (so just 1 channel).

Awesome. Thanks. One last question, is the method applied to your own PPI data exactly the same as Alshahrani,M. et al. (2017)?

The S space is made from the W space. The screenshot says to use any optimization method to get the W+ values. Now, you pass the W+ into the affine...

Hi, sorry for not being clear. I already installed the tensorflow-gpu version. I tested it with pytorch and do see the GPU being used. When I run Magpie, I don't...

Sorry for the late reply. Thanks for this response. I guess the slow part is training w2v. I will do this step separately on a stronger server. Is "easy" to...

In that MIT paper, the lowest sample size is 100. For sample size at 100, BigGAN can probably get pretty good results based on what was reported in SSGAN (the...

I am applying BigGAN on the face dataset https://github.com/apple2373/pytorch-small-dataset-image-generation/tree/master/data from above. I will let you know. One thing to note is that we don't know the best hyperparameters; the readme.md...

I am using my own recent anaconda and pytorch. I manually installed all the libraries required. I found out that using `im.size[0]` and `im.size[1]` returns the height and width. `img.size()`...