search

grooviiee / python_uav

Challenge to Reinforcement learning.
0 stars 0 forks source link

Current issue #21

Closed grooviiee closed 1 year ago

grooviiee commented 1 year ago

UAV agent spits mbs obs shape.

MBS obs.shape: is_uav: False, input_channel 2, input_width 8, input_height 40 hidden_size 64 UAV obs.shape: is_uav: True, input_channel 8, input_width 40, input_height 600 hidden_size 64

RuntimeError: Given groups=1, weight of size [32, 8, 3, 3], expected input[1, 2, 8, 40] to have 8 channels, but got 2 channels instead

grooviiee commented 1 year ago

Issue cleared.

Root cause was in initializing replay buffer I implemented deciding agent is mbs or uav using number of obs.shape.

But it had been changed. And consided every node as mbs.

So, we currently chaged to using is_uav flag