Open rical730 opened 4 years ago
Hello, may I ask how you train the model? When I use "train.py" I can only train "simple" and everything else will give me an error.
Hello, may I ask how you train the model? When I use "train.py" I can only train "simple" and everything else will give me an error.
Hello, have you fixed this issue?
Python (3.5.4) OpenAI gym (0.10.5) tensorflow (1.8.0) numpy (1.14.5)
maddpg code: https://github.com/openai/maddpg
I got an error when I run the maddpg code in the simple scene and display it PS: 8 other environments can work
======================================== Step 1 : train and save model $ python train.py --scenario simple_crypto
2020-01-22 17:46:58.755855: I tensorflow/core/platform/cpu_feature_guard.cc:140] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA Starting iterations... steps: 24975, episodes: 1000, mean episode reward: -22.24953384943575, time: 26.487 steps: 49975, episodes: 2000, mean episode reward: -17.158019479170676, time: 43.373 steps: 74975, episodes: 3000, mean episode reward: -14.317337732475657, time: 43.159 steps: 99975, episodes: 4000, mean episode reward: -7.1827372802086815, time: 45.176
======================================== Step 2 : restore model and display $ python train.py --scenario simple_crypto --display
2020-01-22 17:50:47.412262: I tensorflow/core/platform/cpu_feature_guard.cc:140] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA Loading previous state... Starting iterations... agent 1 to agent 0: B agent 2 to agent 0: C agent 0 to agent 1: A agent 2 to agent 1: C agent 0 to agent 2: A agent 1 to agent 2: B Traceback (most recent call last): File "train.py", line 211, in
train(arglist)
File "train.py", line 170, in train
env.render()
File "/Users/likejiao/Documents/LKJDocument/Code/github/openai/multiagent-particle-envs/multiagent/environment.py", line 234, in render
geom.set_color(*entity.color, alpha=0.5)
TypeError: set_color() got multiple values for argument 'alpha'
======================================== Is there something wrong with the simple_crypto scenario?