diff --git a/tutorials/AgileRL/agilerl_maddpg.py b/tutorials/AgileRL/agilerl_maddpg.py index 902d6125f..14a93f6e1 100644 --- a/tutorials/AgileRL/agilerl_maddpg.py +++ b/tutorials/AgileRL/agilerl_maddpg.py @@ -23,8 +23,8 @@ NET_CONFIG = { "arch": "cnn", # Network architecture "h_size": [32, 32], # Network hidden size - "c_size": [3, 32], # CNN channel size - "k_size": [(1, 3, 3), (1, 3, 3)], # CNN kernel size + "c_size": [32, 32], # CNN channel size + "k_size": [3, 3], # CNN kernel size "s_size": [2, 2], # CNN stride size "normalize": True, # Normalize image from range [0,255] to [0,1] } @@ -36,7 +36,8 @@ # Swap image channels dimension from last to first [H, W, C] -> [C, H, W] "CHANNELS_LAST": True, "BATCH_SIZE": 8, # Batch size - "LR": 0.01, # Learning rate + "LR_ACTOR": 0.001, # Actor learning rate + "LR_CRITIC": 0.01, # Critic learning rate "GAMMA": 0.95, # Discount factor "MEMORY_SIZE": 10000, # Max memory buffer size "LEARN_STEP": 5, # Learning frequency diff --git a/tutorials/AgileRL/agilerl_matd3.py b/tutorials/AgileRL/agilerl_matd3.py index f44db84bb..46aefb5bf 100644 --- a/tutorials/AgileRL/agilerl_matd3.py +++ b/tutorials/AgileRL/agilerl_matd3.py @@ -31,7 +31,8 @@ # Swap image channels dimension from last to first [H, W, C] -> [C, H, W] "CHANNELS_LAST": False, "BATCH_SIZE": 32, # Batch size - "LR": 0.01, # Learning rate + "LR_ACTOR": 0.001, # Actor learning rate + "LR_CRITIC": 0.01, # Critic learning rate "GAMMA": 0.95, # Discount factor "MEMORY_SIZE": 100000, # Max memory buffer size "LEARN_STEP": 5, # Learning frequency diff --git a/tutorials/AgileRL/requirements.txt b/tutorials/AgileRL/requirements.txt index 4a57a7af2..d5eb552e4 100644 --- a/tutorials/AgileRL/requirements.txt +++ b/tutorials/AgileRL/requirements.txt @@ -1,4 +1,5 @@ -agilerl==0.1.19 +agilerl>=0.1.21; python_version >= '3.9' +agilerl==0.1.20; python_version < '3.9' pettingzoo[classic,atari,mpe]>=1.23.1 SuperSuit>=3.9.0 torch>=2.0.1