slm_lab/spec/experimental/dqn/ddqn_lunar.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": 14000
},
"gamma": 0.99,
"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": [
400,
200
],
"hid_layers_activation": "relu",
"clip_grad_val": null,
"loss_spec": {
"name": "MSELoss"
},
"optim_spec": {
"name": "Adam",
"lr": 0.002
},
"lr_scheduler_spec": {
"name": "MultiStepLR",
"milestones": [
80000,
100000
],
"gamma": 0.5
},
"update_type": "polyak",
"polyak_coef": 0.9,
"gpu": false
}
}],
"env": [{
"name": "LunarLander-v2",
"frame_op": "concat",
"frame_op_len": 4,
"max_t": null,
"max_frame": 250000
}],
"body": {
"product": "outer",
"num": 1
},
"meta": {
"distributed": false,
"eval_frequency": 1000,
"max_session": 4,
"max_trial": 62,
"search": "RandomSearch",
},
}
}