slm_lab/spec/experimental/dqn/ddqn_lunar_search.json
{
"ddqn_concat_lunar": {
"agent": [{
"name": "DoubleDQN",
"algorithm": {
"name": "DoubleDQN",
"action_pdtype": "Argmax",
"action_policy": "epsilon_greedy",
"explore_var_spec": {
"name": "linear_decay",
"start_val": 1.0,
"end_val": 0.08,
"start_step": 0,
"end_step": 100000
},
"gamma": 0.99,
"training_batch_iter": 2,
"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": [
400,
200
],
"hid_layers_activation": "relu",
"clip_grad_val": 2.0,
"loss_spec": {
"name": "MSELoss"
},
"optim_spec": {
"name": "Adam",
"lr": 0.001
},
"lr_scheduler_spec": null,
"update_type": "polyak",
"update_frequency": 0.9,
"gpu": false
}
}],
"env": [{
"name": "LunarLander-v2",
"frame_op": "concat",
"frame_op_len": 4,
"max_t": null,
"max_frame": 300000,
"normalize_state": false
}],
"body": {
"product": "outer",
"num": 1
},
"meta": {
"distributed": false,
"log_frequency": 1000,
"eval_frequency": 1000,
"max_session": 4,
"max_trial": 32,
"search": "RandomSearch",
},
"search": {
"agent": [{
"agent": {
"training_batch_iter__randint": [1, 5]
},
"net": {
"update_frequency__choice": [0.8, 0.9, 0.925, 0.95],
"optim_spec": {
"lr__choice": [0.001, 0.003, 0.005, 0.007, 0.01]
}
},
}]
}
}
}