Anh Nguyen

@abadcd How did you solve the ZTIN10tensorflow8OpKernelE problem? I'm encountering the same problem when trying to compile the functions on Google Colab (cuda 10.1, python 3.7, tensorflow 1.15.2)

@Colin97 Do you have more instruction on how to incorporating your implementation in optas' code for latent_3d_points? I'm having issues with compiling the loss functions implemented by optas