yarll.misc.exceptions.ClassNotRegisteredError about yarll HOT 5 CLOSED

Hello,

Thank you for the interest in my code!

Could you tell me how you installed the code and how you ran the experiment?
For me it worked using the latest revision on the tf2 branch by running python yarll/main.py experiment_specs/CartPole-v0-FittedQ-experiment.json.

Note that the version that is installed using pip is quite outdated by now. In case you installed it that way, it is best to clone the repository instead, such that you have the latest version. Also see the README on how to use the code that way.

from yarll.

@arnomoonens

Thank you for your prompt response and for providing this awesome repository. I try to follow your suggestion by installing the cloned version instead. I installed it via pip git+. I have also gone through the README as suggested.

Moreover, after running python yarll/main.py experiment_specs/CartPole-v0-FittedQ-experiment.json, it seems it hanged at : I tensorflow/compiler/mlir/mlir_graph_optimization_pass.cc:176] None of the MLIR Optimization Passes are enabled (registered 2).

Am I to run python -m yarll.misc.plot_statistics <path_to_stats> as the next step?

I anticipate your response.

from yarll.

Update:

So I waited for the experiment to finish running after which I then run python -m yarll.misc.plot_statistics C:\tmp\CartPole-v0-FittedQ, but I got the error below:

line 130, in tf_scalar_data
task = int(pattern.search(run).group(1))
AttributeError: 'NoneType' object has no attribute 'group'

from yarll.

Hello,

There is currently a bug in the plot_statistics file. I suggest to look at statistics using Tensorboard (installed together with TensorFlow). You can run it using tensorboard --logdir C:\tmp\CartPole-v0-FittedQ. You can also run this while the experiment is still running to see if the algorithm is learning well.

from yarll.

Hello,

There is currently a bug in the plot_statistics file. I suggest to look at statistics using Tensorboard (installed together with TensorFlow). You can run it using tensorboard --logdir C:\tmp\CartPole-v0-FittedQ. You can also run this while the experiment is still running to see if the algorithm is learning well.

I will. Thank you!

from yarll.