-
-
https://github.com/facebookresearch/faiss/issues/2531#issuecomment-1280695975 some thoughts here
https://docs.google.com/document/d/1AryWpV0dD_r9x82I_quUzBuRyzDotL_HHnKuNB9H3Zc/edit?usp=drivesdk mo…
-
Hi,
I was trying to understand your model better. In an experiment, I removed the traffic component. I used a dummy linear graph, i.e. nodes are 1,2,3,4...n+1 and a single edge between two consecu…
-
```
~/Documents/pro1/seg/utils/loss_gscnn.py in forward(self, inputs, targets)
344 losses['edge_loss'] = self.edge_weight * 20 * self.bce2d(edgein, edgemask)
345 losses['att…
-
Hi
As per the paper, 4 features must be concatenated as input to TRUNET,
1. log spectrum
2. PCEN
3. real part of demodulated phase
4. imaginary part of demodulated phase
so the input will…
-
- Some key ideas and history behind the EVT, especially focused on Gumbel
- Application of Gumbel softmax to backpropagate through discrete categorical sampling
- Extension of ideas behind Gumbel so…
-
It is possible to construct reparameterizable mixture distributions by replacing the categorical distribution with a gumbel (relaxed categorical) distribution. The ability to use a relaxed one-hot cat…
-
Hi, @tovacinni
Thanks for your work, I have a question about the implementation of dual task loss in loss.py. I think the dual task loss should take the estimation from classical stream N, K, H, W a…
-
This issue is focused on the 4.7.1 version of SC2 and its corresponding replay data. Our first objective is to train a CNN encoder -> LSTM -> CNN decoder network from the data that is able to defeat t…
-
These are the top operators by usage that do not support forward-over-reverse autograd. Supporting these requires writing the forward AD formula for the vjp of the operator. For example, to support fo…