tancik / fourier-feature-networks

Fourier Features Let Networks Learn High Frequency Functions in Low Dimensional Domains
https://people.eecs.berkeley.edu/~bmild/fourfeat/
MIT License
1.26k stars 134 forks source link

Can't run when using neural-tangents 0.3.2 #5

Open EdgeLLM opened 4 years ago

EdgeLLM commented 4 years ago

When I use neural-tangents 0.3.2 to run the 1d_regression.ipynb, I got ValuError: image

Could you help me with this?

tancik commented 4 years ago

Yep, it looks like the neural_tangents 0.3.2 has some breaking changes. I updated the notebook to use version 0.2.2.

fengzlj commented 3 years ago

Hello, When I use neural-tangents 0.2.2 to run the 1d_regression.ipynb, I got AttributeError: 'ShapedArray' object has no attribute 'val': image Could you help me with this?

ShineChen1024 commented 3 years ago

Hello, When I use neural-tangents 0.2.2 to run the 1d_regression.ipynb, I got AttributeError: 'ShapedArray' object has no attribute 'val': image Could you help me with this?

you should try to use jax==0.1.76

geyang commented 3 years ago

@chenshine1 when I use jax==0.1.76 I get the following error

<ipython-input-22-f42eedd5376b> in <module>()
----> 1 import jax
      2 from jax import random, grad, jit, vmap
      3 from jax.config import config
      4 from jax.lib import xla_bridge
      5 import jax.numpy as np

4 frames
/usr/local/lib/python3.7/dist-packages/jax/lib/__init__.py in <module>()
     58   pytree = xla_client._xla.pytree
     59   jax_jit = xla_client._xla.jax_jit
---> 60 from jaxlib import cusolver
     61 try:
     62   from jaxlib import cuda_prng

ImportError: cannot import name 'cusolver' from 'jaxlib' (/usr/local/lib/python3.7/dist-packages/jaxlib/__init__.py)