AlexanderMath

In the demo it takes 1 iteration a second and has 2-4% GPU utilization. 

If you download the pretrained CIFAR10 model it contains several log files 'train.txt', 'test.txt' and 'print.txt'. In 'train.txt' it prints 'train_time' during training, at the last epoch it says ```train_time":...

> Where do you get the pretrained models? It's on the front page, just try search for "pretrain". 

I got the following to work after a few attempts ``` conda create --name keras python -m pip install keras==1.2.2 python -m pip install tensorflow==1.13.0-rc1 ``` I'm pretty convinced these...

Furthermore, if one use an updated version of Pickle there is an error when loading the histopathology dataset. I found the following change fixes the issue: ``` data = pickle.load(f)...

Thanks for the fast reply! The colab would be awesome. I'm not sure what the benefits of docker would be compared to colab.

Thanks for doing this, looks like you managed to get to the bottom of a tricky problem! I'll try to verify at some point, a bit busy atm. If what...

Found same issue with `jax.cumprod` when trying to use log-tricks as temporary solution.

Here's a hacky temporary solution. It uses matrix multiplication to compute `jnp.cumsum` of `2**7` chunks, and then subsequently adds the correct offsets. Use with caution. >90% of time is spent...

Sorry for not writing, already tired what you suggested without luck. Fortunately, the following fixed the issue in my case. I read through the ```environment.yml``` file and realized it doesn't...