Multi_Agent_Deep_Reinforcement_Learning
Multi_Agent_Deep_Reinforcement_Learning copied to clipboard
Problem with converting CNN input
File "/home/lukas/Desktop/Multi_Agent_Deep_Reinforcement_Learning-master/GridWorld_Game_Formulation_Final.py", line 65, in loadModel
ValueError: Negative dimension size caused by subtracting 3 from 2 for 'conv2d_2/convolution' (op: 'Conv2D') with input shapes: [?,2,8,32], [3,3,32,32].
Changing the following line in the "GridWorld_Game_Formulation_Final.ipynb" file work for me.
Old : model.add(Convolution2D(32, 3, 3, input_shape=(4, env.grid_size, env.grid_size),activation='relu'))
New : model.add(Convolution2D(32, 3, 3, input_shape=(env.grid_size, env.grid_size, 4),activation='relu'))