Open Namnodorel opened 1 year ago
I found that the code is not compatible with tensorflow==2.6.0, and one alternative way is to use:
pip install tensorflow==2.4.2 tensorflow_probability==0.12.2
and make sure your python version <= 3.8.
Moreover be cautious about the installation of cuda, if your cuda version is >= 11.0, make sure that you create a soft link for libcusolver.so.10 at the path where your tensorflow lies, for instance for cuda 11.x.x:
ln -s /usr/local/your-cuda/lib64/libcusolver.so.11 ~/path-to-your-python/site-packages/tensorflow/libcusolver.so.10
I wanted to try out dreamerv2 on our own environment (or at least the examples) but unfortunately run into some issues along the way.
The README example & Dockerfile use TensorFlow (and other library) versions that are outdated, in some cases pip doesn't even distribute the older versions anymore.
I attempted to run the minigrid example with a recent TensorFlow version.
The import
from tensorflow.keras.mixed_precision import experimental as prec
innets.py
causes an error as that API is no longer listed under experimental.MiniGrid has migrated elsewhere and now uses Gymnasium: https://github.com/Farama-Foundation/Minigrid
Using the new minigrid env throws an error:
Attempting to call
train.py
with our own environment (with an observation of an intensity image of size 42x30, contained in a NumPy array) manages to collect a prefill dataset but then fails with the errorI was unable to figure out what exactly the problem is here, and whether it is caused by the updated dependency versions, a problem with our environment or something else entirely.
Could you update the dependencies and examples to make it possible again to try out dreamerv2?