Skip to content

viberl.cli

Command-line interface for Tiny RL.

Functions:

Name Description
train_main

Main training CLI entry point.
eval_main

Main evaluation CLI entry point.
demo_main

Demo CLI entry point.

train_main 

train_main()

Main training CLI entry point.

Source code in viberl/cli.py 
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
def train_main():
    """Main training CLI entry point."""
    parser = argparse.ArgumentParser(description='Train RL agents')
    parser.add_argument('--env', choices=['snake'], default='snake', help='Environment to train on')
    parser.add_argument(
        '--alg',
        choices=['reinforce', 'dqn', 'ppo'],
        default='reinforce',
        help='Reinforcement learning algorithm',
    )
    parser.add_argument('--episodes', type=int, default=1000, help='Number of training episodes')
    parser.add_argument('--grid-size', type=int, default=15, help='Grid size for snake environment')
    parser.add_argument('--lr', type=float, default=1e-3, help='Learning rate')
    parser.add_argument('--gamma', type=float, default=0.99, help='Discount factor')
    parser.add_argument('--hidden-size', type=int, default=128, help='Hidden layer size')
    parser.add_argument('--num-hidden-layers', type=int, default=2, help='Number of hidden layers')
    parser.add_argument('--seed', type=int, default=42, help='Random seed')
    parser.add_argument('--name', type=str, help='Experiment name (auto-generated if not provided)')
    parser.add_argument('--render-interval', type=int, help='Render every N episodes')
    parser.add_argument('--save-interval', type=int, help='Save model every N episodes (optional)')
    parser.add_argument('--device', choices=['cpu', 'cuda'], default='auto', help='Device to use')
    parser.add_argument(
        '--epsilon-start', type=float, default=1.0, help='Initial exploration rate (DQN)'
    )
    parser.add_argument(
        '--epsilon-end', type=float, default=0.01, help='Final exploration rate (DQN)'
    )
    parser.add_argument(
        '--epsilon-decay', type=float, default=0.995, help='Exploration decay rate (DQN)'
    )
    parser.add_argument('--memory-size', type=int, default=10000, help='Replay memory size (DQN)')
    parser.add_argument('--batch-size', type=int, default=64, help='Batch size for training')
    parser.add_argument(
        '--target-update', type=int, default=10, help='Target network update frequency (DQN)'
    )
    parser.add_argument('--clip-epsilon', type=float, default=0.2, help='PPO clipping parameter')
    parser.add_argument('--ppo-epochs', type=int, default=4, help='PPO epochs per update')
    parser.add_argument('--gae-lambda', type=float, default=0.95, help='GAE lambda parameter (PPO)')
    parser.add_argument(
        '--value-loss-coef', type=float, default=0.5, help='Value loss coefficient (PPO)'
    )
    parser.add_argument(
        '--entropy-coef', type=float, default=0.01, help='Entropy coefficient (PPO)'
    )
    parser.add_argument('--max-grad-norm', type=float, default=0.5, help='Max gradient norm (PPO)')
    parser.add_argument('--max-steps', type=int, default=1000, help='Maximum steps per episode')
    parser.add_argument('--eval-episodes', type=int, default=10, help='Evaluation episodes')
    parser.add_argument(
        '--eval-interval', type=int, default=100, help='Evaluation interval during training'
    )
    parser.add_argument(
        '--log-interval', type=int, default=1000, help='Log summary interval during training'
    )
    parser.add_argument(
        '--trajectory-batch',
        type=int,
        default=8,
        help='Number of trajectories to collect per training iteration',
    )
    parser.add_argument('--no-eval', action='store_true', help='Skip evaluation after training')
    parser.add_argument('--quiet', action='store_true', help='Suppress training progress output')
    parser.add_argument('--wandb', action='store_true', help='Enable Weights & Biases logging')

    args = parser.parse_args()

    # Set random seed
    set_seed(args.seed)

    # Get device
    device = get_device() if args.device == 'auto' else torch.device(args.device)

    logger.info(f'Using device: {device}')

    # Create environment
    if args.env == 'snake':
        env = SnakeGameEnv(grid_size=args.grid_size)
        state_size = args.grid_size * args.grid_size
        action_size = 4
    else:
        raise ValueError(f'Unknown environment: {args.env}')

    # Create agent
    base_params = {
        'state_size': state_size,
        'action_size': action_size,
        'learning_rate': args.lr,
        'gamma': args.gamma,
        'hidden_size': args.hidden_size,
        'num_hidden_layers': args.num_hidden_layers,
    }

    if args.alg == 'reinforce':
        agent = REINFORCEAgent(**base_params)
    elif args.alg == 'dqn':
        agent = DQNAgent(
            **base_params,
            epsilon_start=args.epsilon_start,
            epsilon_end=args.epsilon_end,
            epsilon_decay=args.epsilon_decay,
            memory_size=args.memory_size,
            batch_size=args.batch_size,
            target_update=args.target_update,
        )
    elif args.alg == 'ppo':
        agent = PPOAgent(
            **base_params,
            clip_epsilon=args.clip_epsilon,
            ppo_epochs=args.ppo_epochs,
            lam=args.gae_lambda,
            value_loss_coef=args.value_loss_coef,
            entropy_coef=args.entropy_coef,
            max_grad_norm=args.max_grad_norm,
            batch_size=args.batch_size,
        )
    else:
        raise ValueError(f'Unknown algorithm: {args.alg}')

    # Move agent to device
    if hasattr(agent, 'policy_network'):
        agent.policy_network.to(device)
    if hasattr(agent, 'q_network'):
        agent.q_network.to(device)
        agent.target_network.to(device)
    if hasattr(agent, 'value_network'):
        agent.value_network.to(device)

    logger.info(f'Training {args.alg} agent on {args.env} environment...')
    logger.info(f'Episodes: {args.episodes}')
    logger.info(f'Grid size: {args.grid_size}')
    logger.info(f'Learning rate: {args.lr}')
    logger.info(f'Gamma: {args.gamma}')

    # Create experiment with automatic directory structure
    experiment_name = args.name or f'{args.alg}_{args.env}'
    exp_manager = create_experiment(experiment_name)
    tb_logs_dir = str(exp_manager.get_tb_logs_path())

    # Configure file logging and log command line arguments
    exp_manager.configure_file_logging(log_level='INFO')
    exp_manager.log_command_line_args(args)

    if tb_logs_dir:
        logger.info(f'TensorBoard logs: {tb_logs_dir}')

    # Create trainer and train agent
    trainer = Trainer(
        env=env,
        agent=agent,
        max_steps=args.max_steps,
        log_dir=tb_logs_dir,
        device=device,
        enable_wandb=args.wandb,
        wandb_config=vars(args),
        run_name=experiment_name,
        batch_size=args.trajectory_batch,
    )

    trainer.train(
        num_episodes=args.episodes,
        render_interval=args.render_interval,
        save_interval=args.save_interval,
        save_path=str(exp_manager.get_models_path()),
        eval_interval=args.eval_interval,
        eval_episodes=args.eval_episodes,
        log_interval=args.log_interval,
        verbose=not args.quiet,
    )

    # Save final model
    models_dir = exp_manager.get_models_path()
    final_model_path = str(models_dir / 'final_model.pth')
    agent.save(final_model_path)
    logger.success(f'Final model saved to {final_model_path}')
    logger.success(f'All experiment files saved in: {exp_manager.get_experiment_path()}')

    env.close()

eval_main 

eval_main()

Main evaluation CLI entry point.

Source code in viberl/cli.py 
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
def eval_main():
    """Main evaluation CLI entry point."""
    parser = argparse.ArgumentParser(description='Evaluate trained RL agents')
    parser.add_argument(
        '--env', choices=['snake'], default='snake', help='Environment to evaluate on'
    )
    parser.add_argument(
        '--agent', choices=['reinforce'], default='reinforce', help='Agent algorithm'
    )
    parser.add_argument('--model-path', type=str, required=True, help='Path to trained model')
    parser.add_argument('--episodes', type=int, default=10, help='Number of evaluation episodes')
    parser.add_argument('--grid-size', type=int, default=15, help='Grid size for snake environment')
    parser.add_argument('--render', action='store_true', help='Render the environment')
    parser.add_argument('--seed', type=int, default=42, help='Random seed')
    parser.add_argument('--device', choices=['cpu', 'cuda'], default='auto', help='Device to use')

    args = parser.parse_args()

    # Set random seed
    set_seed(args.seed)

    # Get device
    device = get_device() if args.device == 'auto' else torch.device(args.device)

    logger.info(f'Using device: {device}')

    # Create environment
    if args.env == 'snake':
        env = SnakeGameEnv(render_mode='human' if args.render else None, grid_size=args.grid_size)
        state_size = args.grid_size * args.grid_size
        action_size = 4
    else:
        raise ValueError(f'Unknown environment: {args.env}')

    # Create agent
    if args.agent == 'reinforce':
        agent = REINFORCEAgent(state_size=state_size, action_size=action_size)
    else:
        raise ValueError(f'Unknown agent: {args.agent}')

    # Move agent to device
    agent.policy_network.to(device)

    # Load trained model
    try:
        agent.load(args.model_path)
        logger.success(f'Loaded model from {args.model_path}')
    except OSError as e:
        logger.error(f'Failed to load model: {e}')
        return

    # Evaluate agent
    logger.info(f'Evaluating {args.agent} agent on {args.env} environment...')

    scores = evaluate_agent(env=env, agent=agent, num_episodes=args.episodes, render=args.render)

    logger.info('\nEvaluation Results:')
    logger.success(f'Average score: {np.mean(scores):.2f} ± {np.std(scores):.2f}')
    logger.info(f'Min score: {np.min(scores):.2f}')
    logger.info(f'Max score: {np.max(scores):.2f}')

    env.close()

demo_main 

demo_main()

Demo CLI entry point.

Source code in viberl/cli.py 
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
def demo_main():
    """Demo CLI entry point."""
    parser = argparse.ArgumentParser(description='Run RL framework demos')
    parser.add_argument('--env', choices=['snake'], default='snake', help='Environment to demo')
    parser.add_argument('--episodes', type=int, default=5, help='Number of demo episodes')
    parser.add_argument('--grid-size', type=int, default=15, help='Grid size for snake environment')
    parser.add_argument('--seed', type=int, default=42, help='Random seed')

    args = parser.parse_args()

    # Set random seed
    set_seed(args.seed)

    # Create environment
    if args.env == 'snake':
        env = SnakeGameEnv(render_mode='human', grid_size=args.grid_size)
    else:
        raise ValueError(f'Unknown environment: {args.env}')

    logger.info(f'Running {args.env} demo for {args.episodes} episodes...')

    # Run demo with random actions
    for episode in range(args.episodes):
        state, info = env.reset()
        total_reward = 0
        steps = 0

        logger.info(f'\nEpisode {episode + 1}/{args.episodes}')

        while True:
            action = env.action_space.sample()  # Random action
            state, reward, terminated, truncated, info = env.step(action)
            total_reward += reward
            steps += 1

            if terminated or truncated:
                logger.success(
                    f'Episode finished! Score: {info.get("score", 0)}, Steps: {steps}, Total reward: {total_reward}'
                )
                break

    env.close()
    logger.success('\nDemo completed!')