slm_lab/spec/benchmark/dqn/dqn_cartpole.json
{
"vanilla_dqn_boltzmann_cartpole": {
"agent": [{
"name": "VanillaDQN",
"algorithm": {
"name": "VanillaDQN",
"action_pdtype": "Categorical",
"action_policy": "boltzmann",
"explore_var_spec": {
"name": "linear_decay",
"start_val": 5.0,
"end_val": 0.5,
"start_step": 0,
"end_step": 4000,
},
"gamma": 0.99,
"training_batch_iter": 8,
"training_iter": 4,
"training_frequency": 4,
"training_start_step": 32
},
"memory": {
"name": "Replay",
"batch_size": 32,
"max_size": 10000,
"use_cer": false
},
"net": {
"type": "MLPNet",
"hid_layers": [64],
"hid_layers_activation": "selu",
"clip_grad_val": 0.5,
"loss_spec": {
"name": "MSELoss"
},
"optim_spec": {
"name": "Adam",
"lr": 0.01
},
"lr_scheduler_spec": {
"name": "LinearToZero",
"frame": 10000
},
"gpu": false
}
}],
"env": [{
"name": "CartPole-v0",
"max_t": null,
"max_frame": 30000
}],
"body": {
"product": "outer",
"num": 1
},
"meta": {
"distributed": false,
"eval_frequency": 500,
"max_session": 4,
"max_trial": 1
},
"search": {
"agent": [{
"net": {
"hid_layers__grid_search": [
[32],
[64],
[32, 16],
[64, 32]
]
}
}]
}
},
"dqn_boltzmann_cartpole": {
"agent": [{
"name": "DQN",
"algorithm": {
"name": "DQN",
"action_pdtype": "Categorical",
"action_policy": "boltzmann",
"explore_var_spec": {
"name": "linear_decay",
"start_val": 5.0,
"end_val": 0.5,
"start_step": 0,
"end_step": 4000,
},
"gamma": 0.99,
"training_batch_iter": 2,
"training_iter": 2,
"training_frequency": 4,
"training_start_step": 32
},
"memory": {
"name": "Replay",
"batch_size": 32,
"max_size": 10000,
"use_cer": false
},
"net": {
"type": "MLPNet",
"hid_layers": [64],
"hid_layers_activation": "selu",
"clip_grad_val": 0.5,
"loss_spec": {
"name": "MSELoss"
},
"optim_spec": {
"name": "Adam",
"lr": 0.02
},
"lr_scheduler_spec": {
"name": "LinearToZero",
"frame": 30000
},
"update_type": "replace",
"update_frequency": 10,
"gpu": false
}
}],
"env": [{
"name": "CartPole-v0",
"max_t": null,
"max_frame": 30000
}],
"body": {
"product": "outer",
"num": 1
},
"meta": {
"distributed": false,
"eval_frequency": 100,
"max_session": 4,
"max_trial": 1
},
"search": {
"agent": [{
"net": {
"hid_layers__grid_search": [
[32],
[64],
[32, 16],
[64, 32]
]
}
}]
}
},
"ddqn_per_boltzmann_cartpole": {
"agent": [{
"name": "DoubleDQN",
"algorithm": {
"name": "DoubleDQN",
"action_pdtype": "Categorical",
"action_policy": "boltzmann",
"explore_var_spec": {
"name": "linear_decay",
"start_val": 5.0,
"end_val": 0.5,
"start_step": 0,
"end_step": 4000,
},
"gamma": 0.99,
"training_batch_iter": 2,
"training_iter": 2,
"training_frequency": 4,
"training_start_step": 32
},
"memory": {
"name": "PrioritizedReplay",
"alpha": 0.6,
"epsilon": 0.0001,
"batch_size": 32,
"max_size": 10000,
"use_cer": false
},
"net": {
"type": "MLPNet",
"hid_layers": [64],
"hid_layers_activation": "selu",
"clip_grad_val": 0.5,
"loss_spec": {
"name": "MSELoss"
},
"optim_spec": {
"name": "Adam",
"lr": 0.005
},
"lr_scheduler_spec": {
"name": "LinearToZero",
"frame": 30000
},
"update_type": "replace",
"update_frequency": 10,
"gpu": false
}
}],
"env": [{
"name": "CartPole-v0",
"max_t": null,
"max_frame": 30000
}],
"body": {
"product": "outer",
"num": 1
},
"meta": {
"distributed": false,
"eval_frequency": 100,
"max_session": 4,
"max_trial": 1
},
"search": {
"agent": [{
"net": {
"hid_layers__grid_search": [
[32],
[64],
[32, 16],
[64, 32]
]
}
}]
}
}
}