Avik Pal

Error in cuda kernels with Pytorch 1.3

3

I am trying to reproduce the model for unsupervised mesh reconstruction (using the same script as in the `examples` directory). But during training I get the following recurring errors (the...

MWE: ```julia using Zygote ps = (weight = randn(Float32, 2, 2),) x = randn(Float32, 2, 1) W = ps.weight # Works gradient(W) do w sum(gradient(x) do y sum(w * y)...

BatchNorm on GPU without affine or tracking statistics

1

Currently, if BatchNorm is being performed on GPU we assert that the parameters must be trainable and statistics must be tracked. Which is a reasonable assumption given CUDNN requires an...

[WIP] Code for training UNet on Cavana Image Masking Kaggle Competition

5

The code is complete and mostly tested. I will remove the WIP tag once I finish training the model. Also will add the Project.toml files and README once training is...

Residual Network

5

Added a sample Residual Network which is going to be trained on CIFAR10. The code can be used as a building block to build deeper Residual Networks. #30

Upsampling GPU Kernel for Flux

5

## Current Issues: - [ ] Line 47 in upsample.jl is not a safe operation without atomic addition. It results in incorrect and inconsistent answers in the backward pass. @maleadt...

Deprecate the Trainer

5

### Background TorchGAN started out with a primary focus on GANs. But recent literature has suggested that GANs are being used in time series modeling, NLP, etc. Hence the name...