bsuite
bsuite copied to clipboard
google-deepmind
bsuite is a collection of carefully-designed experiments that investigate core capabilities of a reinforcement learning (RL) agent
Hi, while working on a PyTorch DQN agent for BSuite experiments, I noticed quite bad results on the mnist and mountain car experiments. I see that a similar question was...
The _total_steps should be increased after the agent takes an action, as in the jax implementation, or during the update function call, as for DQN. If the assignment is done...
It seems like `np.int` has been deprecated in favour of either `int` or `np.int16/32/64/128` [in numpy 1.20](https://numpy.org/devdocs/release/1.20.0-notes.html#using-the-aliases-of-builtin-types-like-np-int-is-deprecated). This PR fixes that error. The exact stack trace fo the error in...
This fixes the `__getattr__` methods in the wrappers to also check if the underlying environments have the attribute called.
`np.int` and analogous expired their deprecation period and have been removed in Numpy 0.24.0: https://numpy.org/doc/stable/release/1.24.0-notes.html#expired-deprecations This causes: ```python import bsuite # Output AttributeError: module 'numpy' has no attribute 'int' ```
Hi, I am observing a strange behavior by the tensorflow default boot dqn agent that I am a bit baffled by. When running sweeps over multiple environments, the agent loses...
I didn't find a parent common interface for all the `bsuite` environments, but a common patter is to have a method `get_observation` to collect the current observation. `Catch`, however, is...
Aligning the `Catch._observation` method to the other environements. See #43 for more.
When trying to run bsuite.environments import catch, I run into the error: ``` AttributeError: module 'numpy' has no attribute 'int'. `np.int` was a deprecated alias for the builtin `int`. To...
BSuite uses `imp` which is deprecated in favour of importlib and removed in 3.12 https://docs.python.org/3.11/library/imp.html
Metadata
Owner
Metadata
bsuite is a collection of carefully-designed experiments that investigate core capabilities of a reinforcement learning (RL) agent