rlgraph / rlgraph

RLgraph: Modular computation graphs for deep reinforcement learning

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Is apex pong broken?

jon-chuang opened this issue · comments

I get the following error when I try running apex_pong.py

File "/home/user/miniconda/envs/py36/lib/python3.6/site-packages/rlgraph/execution/ray/ray_actor.py", line 53, in setup_preprocessor
    ), action_space=None)
File "/home/user/miniconda/envs/py36/lib/python3.6/site-packages/rlgraph/components/layers/preprocessing/grayscale.py", line 73, in create_variables
    in_space = input_spaces["inputs"]
KeyError: 'inputs'

On the other hand, if I change the key to preprocessing_inputs, some section of the code feeds inputs.

This error also occurs with

/home/user/miniconda/envs/py36/lib/python3.6/site-packages/rlgraph/components/layers/preprocessing/image_resize.py

and I suspect every other file in preprocessing

This was fixed for me easily by adding an "inputs" to the fed dict in ray_actor.py

processor_stack.sub_components[sub_comp_scope].create_variables(input_spaces=dict(
                    preprocessing_inputs=build_space, inputs=build_space
                ), action_space=None)
```

Many thanks for the spot, fixed.