Attribute Error: module 'tensorflow_datasets.core' has no attribute 'ReadWritePath'

[google-deepmind]

Attribute Error: module 'tensorflow_datasets.core' has no attribute 'ReadWritePath'

[yangyi02]

It may be because the kubric package you installed is outdated.

Could you remove the kubric from python3.10 site-packages and follow the instructions?

Particularly you don't need to install kubric package, just need to download kubric github codebase as a subdirectory,

Install kubric as a subdirectory:

git clone https://github.com/google-research/kubric.git

Install requirements:

pip install -r requirements.txt

[cdoersch]

Looks like tfds.core.ReadWritePath was removed a long time ago (february of last year; tensorflow/datasets@b25e2a2#diff-f6f15af3a15c244b2fe4273df8917f17debb279d192051d58b84b14568406fe1 ). However, custom_types.py doesn't even exist in recent versions of kubric, so this import shouldn't fail if your kubric is new enough.

However, I do see a couple of references to tfds.core.ReadWritePath still in kubric; shouldn't affect your use case, but I'll ping the kubric team about it.