milmor

What's the best way to fix TF spectral norm? It's been more than a year, and this layer still doesn't work.

> That means that CuDNN is not in your library path. Can you try adding your CUDA lib path to `LD_LIBRARY_PATH`? After ``` $ export PATH=/usr/local/cuda-11.1/bin${PATH:+:${PATH}} $ export LD_LIBRARY_PATH=/usr/local/cuda-11.1/lib64\ ${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}...

> About 5 months ago ([a141cc6](https://github.com/google/jax/commit/a141cc6e8d36ff10e28180683588bedf5432df1a)) we switched how we link GPU libraries to be the same as TensorFlow, namely, we use `dlopen()` to find libraries like CuDNN rather than...

> > Tensorflow 2.3 works perfect, no error. > > Do you mean _upgrading_ to 2.3 or _downgrading_ to 2.3? After creating a new conda enviroment, and installing tensorflow-gpu==2.3 using...

> Looks like this may have been fixed in 0.2.6. @milmor can you confirm? The issue has not been fixed in 0.2.6. I found that although jaxlib .whl is for...

Using the default imagenet 128 x 128 model and model.summary(), the number of parameters is 32M. There is no model with 67M parameters as mentioned in the paper.