A small bug

Open George-Chia opened this issue 4 years ago • 1 comments

I am reading your excellent code. I found a small bug that failing to load the imitation model. I suggest replacing model.load_state_dict_rl(torch.load(il_weight_file)) with policy.load_state_dict_rl(torch.load(il_weight_file)) in line 138 in train.py. Thanks!

Apr 15 '21 14:04 George-Chia

It looks like by default the policy saves the weights of policy.model https://github.com/ChanganVR/RelationalGraphLearning/blob/8e87aa5ed8221efd688f8e6857ba4c38637bf6e1/crowd_sim/envs/policy/policy.py#L39-L40

Only in model-predictive RL, it saves the weights differently. Were you trying to load the weights of model-predictive RL as initialization?

Apr 22 '21 20:04 ChanganVR