kengz/SLM-Lab

View on GitHub
slm_lab/spec/demo.json

Summary

Maintainability
Test Coverage
{
  "dqn_cartpole": {
    "agent": [{
      "name": "DQN",
      "algorithm": {
        "name": "DQN",
        "action_pdtype": "Argmax",
        "action_policy": "epsilon_greedy",
        "explore_var_spec": {
          "name": "linear_decay",
          "start_val": 1.0,
          "end_val": 0.1,
          "start_step": 0,
          "end_step": 1000
        },
        "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": true
      },
      "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": "StepLR",
          "step_size": 1000,
          "gamma": 0.9
        },
        "update_type": "polyak",
        "update_frequency": 32,
        "polyak_coef": 0.1,
        "gpu": false
      }
    }],
    "env": [{
      "name": "CartPole-v0",
      "max_t": null,
      "max_frame": 10000
    }],
    "body": {
      "product": "outer",
      "num": 1
    },
    "meta": {
      "distributed": false,
      "eval_frequency": 500,
      "log_frequency": 500,
      "max_session": 2,
      "max_trial": 1
    },
    "search": {
      "agent": [{
        "algorithm": {
          "gamma__grid_search": [0.5, 0.7, 0.90, 0.95, 0.99]
        }
      }]
    }
  }
}