benchmark(davide): Bsuite memory benchmark (#138)

* added r2d2 + a2c configs

* changed convergence reward for some env

* removed configs that don't converge

* removed 'on_policy' param in 2rd2 configs

dizoo/bsuite/config/serial/memory_len/memory_len_0_a2c_config.py

+from easydict import EasyDict
+from ding.entry import serial_pipeline
+
+memory_len_a2c_config = dict(
+    exp_name='memory_len_0_a2c',
+    env=dict(
+        collector_env_num=8,
+        evaluator_env_num=1,
+        n_evaluator_episode=100,
+        env_id='memory_len/0',
+        stop_value=1.,
+    ),
+    policy=dict(
+        cuda=False,
+        # (bool) whether use on-policy training pipeline(behaviour policy and training policy are the same)
+        model=dict(
+            obs_shape=3,
+            action_shape=2,
+            encoder_hidden_size_list=[128, 128, 64],
+        ),
+        learn=dict(
+            batch_size=64,
+            # (bool) Whether to normalize advantage. Default to False.
+            normalize_advantage=False,
+            learning_rate=0.001,
+            # (float) loss weight of the value network, the weight of policy network is set to 1
+            value_weight=0.5,
+            # (float) loss weight of the entropy regularization, the weight of policy network is set to 1
+            entropy_weight=0.01,
+        ),
+        collect=dict(
+            # (int) collect n_sample data, train model n_iteration times
+            n_sample=80,
+            # (float) the trade-off factor lambda to balance 1step td and mc
+            gae_lambda=0.95,
+        ),
+        eval=dict(evaluator=dict(eval_freq=100, )),
+    ),
+)
+memory_len_a2c_config = EasyDict(memory_len_a2c_config)
+main_config = memory_len_a2c_config
+
+memory_len_a2c_create_config = dict(
+    env=dict(
+        type='bsuite',
+        import_names=['dizoo.bsuite.envs.bsuite_env'],
+    ),
+    env_manager=dict(type='base'),
+    policy=dict(type='a2c'),
+)
+memory_len_a2c_create_config = EasyDict(memory_len_a2c_create_config)
+create_config = memory_len_a2c_create_config
+
+if __name__ == "__main__":
+    serial_pipeline([main_config, create_config], seed=0)
\ No newline at end of file

dizoo/bsuite/config/serial/memory_len/memory_len_0_dqn_config.py

 from easydict import EasyDict
+from ding.entry import serial_pipeline
 
-memory_len_0_dqn_config = dict(
+memory_len_dqn_config = dict(
     exp_name='memory_len_0_dqn',
     env=dict(
         collector_env_num=8,
         evaluator_env_num=1,
-        n_evaluator_episode=10,
+        n_evaluator_episode=100,
         env_id='memory_len/0',
         stop_value=1.,
     ),
@@ -25,7 +26,7 @@ memory_len_0_dqn_config = dict(
             learning_rate=0.001,
         ),
         collect=dict(n_sample=8),
-        eval=dict(evaluator=dict(eval_freq=20, )),
+        eval=dict(evaluator=dict(eval_freq=100, )),
         other=dict(
             eps=dict(
                 type='exp',
@@ -37,9 +38,9 @@ memory_len_0_dqn_config = dict(
         ),
     ),
 )
-memory_len_0_dqn_config = EasyDict(memory_len_0_dqn_config)
-main_config = memory_len_0_dqn_config
-memory_len_0_dqn_create_config = dict(
+memory_len_dqn_config = EasyDict(memory_len_dqn_config)
+main_config = memory_len_dqn_config
+memory_len_dqn_create_config = dict(
     env=dict(
         type='bsuite',
         import_names=['dizoo.bsuite.envs.bsuite_env'],
@@ -47,5 +48,8 @@ memory_len_0_dqn_create_config = dict(
     env_manager=dict(type='base'),
     policy=dict(type='dqn'),
 )
-memory_len_0_dqn_create_config = EasyDict(memory_len_0_dqn_create_config)
-create_config = memory_len_0_dqn_create_config
+memory_len_dqn_create_config = EasyDict(memory_len_dqn_create_config)
+create_config = memory_len_dqn_create_config
+
+if __name__ == "__main__":
+    serial_pipeline([main_config, create_config], seed=0)

dizoo/bsuite/config/serial/memory_len/memory_len_10_r2d2_config.py

+from easydict import EasyDict
+from ding.entry import serial_pipeline
+
+collector_env_num = 8
+evaluator_env_num = 1
+memory_len_r2d2_config = dict(
+    exp_name='memory_len_10_r2d2',
+    env=dict(
+        collector_env_num=collector_env_num,
+        evaluator_env_num=evaluator_env_num,
+        n_evaluator_episode=100,
+        env_id='memory_len/10',  # 12 memory steps, 13 obs per episode
+        stop_value=1.,
+    ),
+    policy=dict(
+        cuda=True,
+        priority=True,
+        priority_IS_weight=True,
+        model=dict(
+            obs_shape=3,
+            action_shape=2,
+            encoder_hidden_size_list=[128, 128, 64],
+        ),
+        discount_factor=0.997,
+        burnin_step=1,
+        nstep=2,
+        # (int) the whole sequence length to unroll the RNN network minus
+        # the timesteps of burnin part,
+        # i.e., <the whole sequence length> = <burnin_step> + <unroll_len>
+        unroll_len=15,
+        learn=dict(
+            # according to the R2D2 paper, actor parameter update interval is 400
+            # environment timesteps, and in per collect phase, we collect 32 sequence
+            # samples, the length of each sample sequence is <burnin_step> + <unroll_len>,
+            # which is 100 in our seeting, 32*100/400=8, so we set update_per_collect=8
+            # in most environments
+            update_per_collect=8,
+            batch_size=64,
+            learning_rate=0.0005,
+            target_update_theta=0.001,
+        ),
+        collect=dict(
+            # NOTE it is important that don't include key n_sample here, to make sure self._traj_len=INF
+            each_iter_n_sample=32,
+            env_num=collector_env_num,
+        ),
+        eval=dict(env_num=evaluator_env_num, evaluator=dict(eval_freq=100, )),
+        other=dict(
+            eps=dict(
+                type='exp',
+                start=0.95,
+                end=0.05,
+                decay=1e5,
+            ),
+            replay_buffer=dict(replay_buffer_size=50000,
+                               # (Float type) How much prioritization is used: 0 means no prioritization while 1 means full prioritization
+                               alpha=0.6,
+                               # (Float type)  How much correction is used: 0 means no correction while 1 means full correction
+                               beta=0.4,
+                               )
+        ),
+    ),
+)
+memory_len_r2d2_config = EasyDict(memory_len_r2d2_config)
+main_config = memory_len_r2d2_config
+memory_len_r2d2_create_config = dict(
+    env=dict(
+        type='bsuite',
+        import_names=['dizoo.bsuite.envs.bsuite_env'],
+    ),
+    env_manager=dict(type='base'),
+    policy=dict(type='r2d2'),
+)
+memory_len_r2d2_create_config = EasyDict(memory_len_r2d2_create_config)
+create_config = memory_len_r2d2_create_config
+
+if __name__ == "__main__":
+    serial_pipeline([main_config, create_config], seed=0)

dizoo/bsuite/config/serial/memory_len/memory_len_17_r2d2_config.py

+from easydict import EasyDict
+from ding.entry import serial_pipeline
+
+collector_env_num = 8
+evaluator_env_num = 1
+memory_len_r2d2_config = dict(
+    exp_name='memory_len_17_r2d2',
+    env=dict(
+        collector_env_num=collector_env_num,
+        evaluator_env_num=evaluator_env_num,
+        n_evaluator_episode=100,
+        env_id='memory_len/17',  # 50 memory steps, 51 obs per episode
+        stop_value=1.,
+    ),
+    policy=dict(
+        cuda=True,
+        priority=True,
+        priority_IS_weight=True,
+        model=dict(
+            obs_shape=3,
+            action_shape=2,
+            encoder_hidden_size_list=[128, 128, 64],
+        ),
+        discount_factor=0.997,
+        burnin_step=1,
+        nstep=2,
+        # (int) the whole sequence length to unroll the RNN network minus
+        # the timesteps of burnin part,
+        # i.e., <the whole sequence length> = <burnin_step> + <unroll_len>
+        unroll_len=60,
+        learn=dict(
+            # according to the R2D2 paper, actor parameter update interval is 400
+            # environment timesteps, and in per collect phase, we collect 32 sequence
+            # samples, the length of each sample sequence is <burnin_step> + <unroll_len>,
+            # which is 100 in our seeting, 32*100/400=8, so we set update_per_collect=8
+            # in most environments
+            update_per_collect=8,
+            batch_size=64,
+            learning_rate=0.0005,
+            target_update_theta=0.001,
+        ),
+        collect=dict(
+            # NOTE it is important that don't include key n_sample here, to make sure self._traj_len=INF
+            each_iter_n_sample=32,
+            env_num=collector_env_num,
+        ),
+        eval=dict(env_num=evaluator_env_num, evaluator=dict(eval_freq=100, )),
+        other=dict(
+            eps=dict(
+                type='exp',
+                start=0.95,
+                end=0.05,
+                decay=1e5,
+            ),
+            replay_buffer=dict(replay_buffer_size=50000,
+                               # (Float type) How much prioritization is used: 0 means no prioritization while 1 means full prioritization
+                               alpha=0.6,
+                               # (Float type)  How much correction is used: 0 means no correction while 1 means full correction
+                               beta=0.4,
+                               )
+        ),
+    ),
+)
+memory_len_r2d2_config = EasyDict(memory_len_r2d2_config)
+main_config = memory_len_r2d2_config
+memory_len_r2d2_create_config = dict(
+    env=dict(
+        type='bsuite',
+        import_names=['dizoo.bsuite.envs.bsuite_env'],
+    ),
+    env_manager=dict(type='base'),
+    policy=dict(type='r2d2'),
+)
+memory_len_r2d2_create_config = EasyDict(memory_len_r2d2_create_config)
+create_config = memory_len_r2d2_create_config
+
+if __name__ == "__main__":
+    serial_pipeline([main_config, create_config], seed=0)