slm_lab/spec/experimental/misc/gridworld.json
{
"a2c_gae_mlp_separate_gridworld": {
"agent": [{
"name": "A2C",
"algorithm": {
"name": "ActorCritic",
"action_pdtype": "default",
"action_policy": "default",
"explore_var_spec": null,
"gamma": 0.999,
"lam": 0.97,
"num_step_returns": null,
"entropy_coef_spec": null,
"policy_loss_coef": 1.0,
"val_loss_coef": 1.0,
"training_frequency": 1,
},
"memory": {
"name": "OnPolicyReplay"
},
"net": {
"type": "MLPNet",
"shared": false,
"hid_layers": [200],
"hid_layers_activation": "sigmoid",
"clip_grad_val": null,
"use_same_optim": false,
"actor_optim_spec": {
"name": "Adam",
"lr": 0.01
},
"critic_optim_spec": {
"name": "Adam",
"lr": 0.01
},
"lr_scheduler_spec": {
"name": "StepLR",
"step_size": 500,
"gamma": 0.9,
},
"gpu": false
}
}],
"env": [{
"name": "gridworld",
"max_t": null,
"max_frame": 1000,
}],
"body": {
"product": "outer",
"num": 1
},
"meta": {
"distributed": false,
"eval_frequency": 1000,
"max_session": 4,
"max_trial": 95,
"search": "RandomSearch",
},
"search": {
"agent": [{
"algorithm": {
"training_frequency__choice": [1, 2, 3],
"lam__uniform": [0.9, 0.99]
},
"net": {
"hid_layers_activation__choice": ["relu", "sigmoid"],
"lr_decay_frequency__choice": [30000, 40000, 50000, 60000, 70000],
"actor_optim_spec": {
"lr__uniform": [0.001, 0.01]
},
"critic_optim_spec": {
"lr__uniform": [0.001, 0.01]
}
}
}]
}
},
"a2c_gae_rnn_separate_gridworld": {
"agent": [{
"name": "A2C",
"algorithm": {
"name": "ActorCritic",
"action_pdtype": "default",
"action_policy": "default",
"explore_var_spec": null,
"gamma": 0.999,
"lam": 0.97,
"num_step_returns": null,
"entropy_coef_spec": null,
"policy_loss_coef": 1.0,
"val_loss_coef": 1.0,
"training_frequency": 1,
},
"memory": {
"name": "OnPolicyReplay"
},
"net": {
"type": "RecurrentNet",
"shared": false,
"cell_type": "GRU",
"fc_hid_layers": [200],
"hid_layers_activation": "sigmoid",
"rnn_hidden_size": 64,
"rnn_num_layers": 1,
"seq_len": 4,
"clip_grad_val": null,
"use_same_optim": false,
"actor_optim_spec": {
"name": "Adam",
"lr": 0.01
},
"critic_optim_spec": {
"name": "Adam",
"lr": 0.01
},
"lr_scheduler_spec": {
"name": "StepLR",
"step_size": 500,
"gamma": 0.9,
},
"gpu": false
}
}],
"env": [{
"name": "gridworld",
"max_t": null,
"max_frame": 1000,
}],
"body": {
"product": "outer",
"num": 1
},
"meta": {
"distributed": false,
"eval_frequency": 1000,
"max_session": 4,
"max_trial": 95,
"search": "RandomSearch",
},
"search": {
"agent": [{
"algorithm": {
"training_frequency__choice": [1, 2, 3],
"lam__uniform": [0.9, 0.99]
},
"net": {
"hid_layers_activation__choice": ["relu", "sigmoid"],
"lr_decay_frequency__choice": [30000, 40000, 50000, 60000, 70000],
"actor_optim_spec": {
"lr__uniform": [0.001, 0.01]
},
"critic_optim_spec": {
"lr__uniform": [0.001, 0.01]
}
}
}]
}
},
"a2c_nstep_mlp_separate_gridworld": {
"agent": [{
"name": "A2C",
"algorithm": {
"name": "ActorCritic",
"action_pdtype": "default",
"action_policy": "default",
"explore_var_spec": null,
"gamma": 0.999,
"lam": null,
"num_step_returns": 1,
"entropy_coef_spec": null,
"policy_loss_coef": 1.0,
"val_loss_coef": 1.0,
"training_frequency": 1,
},
"memory": {
"name": "OnPolicyReplay"
},
"net": {
"type": "MLPNet",
"shared": false,
"hid_layers": [200],
"hid_layers_activation": "sigmoid",
"clip_grad_val": null,
"use_same_optim": false,
"actor_optim_spec": {
"name": "Adam",
"lr": 0.01
},
"critic_optim_spec": {
"name": "Adam",
"lr": 0.01
},
"lr_scheduler_spec": {
"name": "StepLR",
"step_size": 500,
"gamma": 0.9,
},
"gpu": false
}
}],
"env": [{
"name": "gridworld",
"max_t": null,
"max_frame": 1000,
}],
"body": {
"product": "outer",
"num": 1
},
"meta": {
"distributed": false,
"eval_frequency": 1000,
"max_session": 4,
"max_trial": 95,
"search": "RandomSearch",
},
"search": {
"agent": [{
"algorithm": {
"training_frequency__choice": [1, 2, 3],
"num_step_returns__choice": [1, 2, 3, 5, 10]
},
"net": {
"hid_layers_activation__choice": ["relu", "sigmoid"],
"lr_decay_frequency__choice": [30000, 40000, 50000, 60000, 70000],
"actor_optim_spec": {
"lr__uniform": [0.001, 0.01]
},
"critic_optim_spec": {
"lr__uniform": [0.001, 0.01]
}
}
}]
}
},
"a2c_nstep_rnn_separate_gridworld": {
"agent": [{
"name": "A2C",
"algorithm": {
"name": "ActorCritic",
"action_pdtype": "default",
"action_policy": "default",
"explore_var_spec": null,
"gamma": 0.999,
"lam": null,
"num_step_returns": 1,
"entropy_coef_spec": null,
"policy_loss_coef": 1.0,
"val_loss_coef": 1.0,
"training_frequency": 1,
},
"memory": {
"name": "OnPolicyReplay"
},
"net": {
"type": "RecurrentNet",
"shared": false,
"cell_type": "GRU",
"fc_hid_layers": [200],
"hid_layers_activation": "sigmoid",
"rnn_hidden_size": 64,
"rnn_num_layers": 1,
"seq_len": 4,
"clip_grad_val": null,
"use_same_optim": false,
"actor_optim_spec": {
"name": "Adam",
"lr": 0.01
},
"critic_optim_spec": {
"name": "Adam",
"lr": 0.01
},
"lr_scheduler_spec": {
"name": "StepLR",
"step_size": 500,
"gamma": 0.9,
},
"gpu": false
}
}],
"env": [{
"name": "gridworld",
"max_t": null,
"max_frame": 1000,
}],
"body": {
"product": "outer",
"num": 1
},
"meta": {
"distributed": false,
"eval_frequency": 1000,
"max_session": 4,
"max_trial": 95,
"search": "RandomSearch",
},
"search": {
"agent": [{
"algorithm": {
"training_frequency__choice": [1, 2, 3],
"num_step_returns__choice": [1, 2, 3, 5, 10]
},
"net": {
"hid_layers_activation__choice": ["relu", "sigmoid"],
"lr_decay_frequency__choice": [30000, 40000, 50000, 60000, 70000],
"actor_optim_spec": {
"lr__uniform": [0.001, 0.01]
},
"critic_optim_spec": {
"lr__uniform": [0.001, 0.01]
}
}
}]
}
},
"dqn_boltzmann_gridworld": {
"agent": [{
"name": "DQN",
"algorithm": {
"name": "DQN",
"action_pdtype": "Categorical",
"action_policy": "boltzmann",
"explore_var_spec": {
"name": "linear_decay",
"start_val": 3.0,
"end_val": 1.0,
"start_step": 0,
"end_step": 2000,
},
"gamma": 0.999,
"training_batch_iter": 3,
"training_iter": 4,
"training_frequency": 4,
"training_start_step": 32
},
"memory": {
"name": "Replay",
"batch_size": 32,
"max_size": 100000,
"use_cer": false
},
"net": {
"type": "MLPNet",
"hid_layers": [200],
"hid_layers_activation": "sigmoid",
"clip_grad_val": null,
"loss_spec": {
"name": "MSELoss"
},
"optim_spec": {
"name": "Adam",
"lr": 0.01
},
"lr_scheduler_spec": {
"name": "StepLR",
"step_size": 500,
"gamma": 0.9,
},
"update_type": "polyak",
"update_frequency": 200,
"polyak_coef": 0,
"gpu": false
}
}],
"env": [{
"name": "gridworld",
"frame_op": "concat",
"frame_op_len": 4,
"max_t": null,
"max_frame": 1000,
}],
"body": {
"product": "outer",
"num": 1
},
"meta": {
"distributed": false,
"eval_frequency": 1000,
"max_session": 4,
"max_trial": 95,
"search": "RandomSearch",
},
"search": {
"agent": [{
"algorithm": {
"explore_var_spec": {
"end_step__choice": [30000, 40000, 50000]
}
},
"memory": {
"max_size__choice": [10000, 50000, 100000],
},
"net": {
"training_iter__choice": [1, 2, 3, 4],
"lr_decay_frequency__choice": [30000, 40000, 50000, 60000, 70000],
"polyak_coef__choice": [0, 0.9, 0.99, 0.999],
"lr__uniform": [0.001, 0.01],
"hid_layers_activation__choice": ["relu", "sigmoid"],
}
}]
}
},
"drqn_boltzmann_gridworld": {
"agent": [{
"name": "DQN",
"algorithm": {
"name": "DQN",
"action_pdtype": "Categorical",
"action_policy": "boltzmann",
"explore_var_spec": {
"name": "linear_decay",
"start_val": 3.0,
"end_val": 1.0,
"start_step": 0,
"end_step": 2000,
},
"gamma": 0.999,
"training_batch_iter": 3,
"training_iter": 4,
"training_frequency": 4,
"training_start_step": 32
},
"memory": {
"name": "Replay",
"batch_size": 32,
"max_size": 100000,
"use_cer": false
},
"net": {
"type": "RecurrentNet",
"cell_type": "GRU",
"fc_hid_layers": [200],
"hid_layers_activation": "sigmoid",
"rnn_hidden_size": 64,
"rnn_num_layers": 1,
"seq_len": 4,
"clip_grad_val": null,
"loss_spec": {
"name": "MSELoss"
},
"optim_spec": {
"name": "Adam",
"lr": 0.01
},
"lr_scheduler_spec": {
"name": "StepLR",
"step_size": 500,
"gamma": 0.9,
},
"update_type": "polyak",
"update_frequency": 200,
"polyak_coef": 0,
"gpu": false
}
}],
"env": [{
"name": "gridworld",
"frame_op": "concat",
"frame_op_len": 4,
"max_t": null,
"max_frame": 1000,
}],
"body": {
"product": "outer",
"num": 1
},
"meta": {
"distributed": false,
"eval_frequency": 1000,
"max_session": 4,
"max_trial": 95,
"search": "RandomSearch",
},
"search": {
"agent": [{
"algorithm": {
"explore_var_spec": {
"end_step__choice": [30000, 40000, 50000]
}
},
"memory": {
"max_size__choice": [10000, 50000, 100000],
},
"net": {
"rnn_hidden_size__choice": [32, 64],
"seq_len__choice": [2, 3, 4, 5, 6],
"training_iter__choice": [1, 2, 3, 4],
"lr_decay_frequency__choice": [30000, 40000, 50000, 60000, 70000],
"lr__uniform": [0.001, 0.01],
"polyak_coef__choice": [0, 0.9, 0.99, 0.999],
"hid_layers_activation__choice": ["relu", "sigmoid"],
}
}]
}
},
"ddqn_boltzmann_gridworld": {
"agent": [{
"name": "DoubleDQN",
"algorithm": {
"name": "DoubleDQN",
"action_pdtype": "Categorical",
"action_policy": "boltzmann",
"explore_var_spec": {
"name": "linear_decay",
"start_val": 3.0,
"end_val": 1.0,
"start_step": 0,
"end_step": 2000,
},
"gamma": 0.999,
"training_batch_iter": 3,
"training_iter": 4,
"training_frequency": 4,
"training_start_step": 32
},
"memory": {
"name": "Replay",
"batch_size": 32,
"max_size": 100000,
"use_cer": false
},
"net": {
"type": "MLPNet",
"hid_layers": [200],
"hid_layers_activation": "sigmoid",
"clip_grad_val": null,
"loss_spec": {
"name": "MSELoss"
},
"optim_spec": {
"name": "Adam",
"lr": 0.01
},
"lr_scheduler_spec": {
"name": "StepLR",
"step_size": 500,
"gamma": 0.9,
},
"update_type": "polyak",
"update_frequency": 200,
"polyak_coef": 0,
"gpu": false
}
}],
"env": [{
"name": "gridworld",
"frame_op": "concat",
"frame_op_len": 4,
"max_t": null,
"max_frame": 1000,
}],
"body": {
"product": "outer",
"num": 1
},
"meta": {
"distributed": false,
"eval_frequency": 1000,
"max_session": 4,
"max_trial": 95,
"search": "RandomSearch",
},
"search": {
"agent": [{
"algorithm": {
"explore_var_spec": {
"end_step__choice": [30000, 40000, 50000]
}
},
"memory": {
"max_size__choice": [10000, 50000, 100000],
},
"net": {
"training_iter__choice": [1, 2, 3, 4],
"lr_decay_frequency__choice": [30000, 40000, 50000, 60000, 70000],
"lr__uniform": [0.001, 0.01],
"polyak_coef__choice": [0, 0.9, 0.99, 0.999],
"hid_layers_activation__choice": ["relu", "sigmoid"],
}
}]
}
},
"ddrqn_boltzmann_gridworld": {
"agent": [{
"name": "DoubleDQN",
"algorithm": {
"name": "DoubleDQN",
"action_pdtype": "Categorical",
"action_policy": "boltzmann",
"explore_var_spec": {
"name": "linear_decay",
"start_val": 3.0,
"end_val": 1.0,
"start_step": 0,
"end_step": 2000,
},
"gamma": 0.999,
"training_batch_iter": 3,
"training_iter": 4,
"training_frequency": 4,
"training_start_step": 32
},
"memory": {
"name": "Replay",
"batch_size": 32,
"max_size": 100000,
"use_cer": false
},
"net": {
"type": "RecurrentNet",
"cell_type": "GRU",
"fc_hid_layers": [200],
"hid_layers_activation": "sigmoid",
"rnn_hidden_size": 64,
"rnn_num_layers": 1,
"seq_len": 4,
"clip_grad_val": null,
"loss_spec": {
"name": "MSELoss"
},
"optim_spec": {
"name": "Adam",
"lr": 0.01
},
"lr_scheduler_spec": {
"name": "StepLR",
"step_size": 500,
"gamma": 0.9,
},
"update_type": "polyak",
"update_frequency": 200,
"polyak_coef": 0,
"gpu": false
}
}],
"env": [{
"name": "gridworld",
"frame_op": "concat",
"frame_op_len": 4,
"max_t": null,
"max_frame": 1000,
}],
"body": {
"product": "outer",
"num": 1
},
"meta": {
"distributed": false,
"eval_frequency": 1000,
"max_session": 4,
"max_trial": 95,
"search": "RandomSearch",
},
"search": {
"agent": [{
"algorithm": {
"explore_var_spec": {
"end_step__choice": [30000, 40000, 50000]
}
},
"memory": {
"max_size__choice": [10000, 50000, 100000],
},
"net": {
"rnn_hidden_size__choice": [32, 64],
"seq_len__choice": [2, 3, 4, 5, 6],
"training_iter__choice": [1, 2, 3, 4],
"lr_decay_frequency__choice": [30000, 40000, 50000, 60000, 70000],
"lr__uniform": [0.001, 0.01],
"polyak_coef__choice": [0, 0.9, 0.99, 0.999],
"hid_layers_activation__choice": ["relu", "sigmoid"],
}
}]
}
},
}