vincent341

How about changing the shape of feature maps by "x = x.view(xs[0], xs[1] * xs[2], xs[3], xs[4])", then feeding it to BatchNorm2d, where xs is the shape of x?

> train_info Hi Min Zhao, would you mind giving more details about your solution? How did you make it work on Mp3D? It would be very helpful to me. Thanks...

Hi @erikwijmans . I run into the error "Platform::WindowlessEglApplication::tryCreateContext(): unable to find EGL device for CUDA device 0" too when I was working on habitat challenge. I didn't find "libEGL_nvidia.so.0...

> Those are in the normal cuda-drivers package Hi @erikwijmans , Thanks for your quick reply. Does it mean that libEGL_nvidia.so.0 should be in the computer if the cuda driver...

@erikwijmans Thanks very much for your help.

Hi @erikwijmans , I'm still struggling with this problem. Previously, an "unable to fine EGL device for CUDA device 0" occured. At that time, I checked the computer. libEGL_nvidia.so* cannot...

> @vincent341 Hi, I meet the same issue, have you solved this problem? Hi, it has been some time. I'm not sure if I remember correctly. I solved it by...

Hi. I wrote a LNLSTM (https://github.com/vincent341/LayerNormLSTM/blob/master/LayerNormLSTM.py) implementing Layer Normalization(https://arxiv.org/abs/1607.06450). Actually I modified the original LSTM Layer by referencing this code(https://github.com/ryankiros/layer-norm). I want to give great thanks to Jan Schlüter for...

> Hi, > > I'm not sure if this is an issue, or that's my misunderstanding. You calculated loss as: > > loss = class_loss + params.theta * domain_loss >...

> > > Hi, > > > I'm not sure if this is an issue, or that's my misunderstanding. You calculated loss as: > > > loss = class_loss +...