Fix/bugs

sheeprl/algos/ppo/ppo.py

@@ -300,7 +300,7 @@ def main(fabric: Fabric, cfg: Dict[str, Any]):
                         for k, v in info["final_observation"][truncated_env].items():
                             torch_v = torch.as_tensor(v, dtype=torch.float32, device=device)
                             if k in cfg.algo.cnn_keys.encoder:
-                                torch_v = torch_v.view(cfg.env.num_envs, -1, *v.shape[-2:])
+                                torch_v = torch_v.view(-1, *v.shape[-2:])
                                 torch_v = torch_v / 255.0 - 0.5
                             real_next_obs[k][i] = torch_v
                     with torch.no_grad():

sheeprl/algos/ppo/ppo_decoupled.py

@@ -222,7 +222,7 @@ def player(
                         for k, v in info["final_observation"][truncated_env].items():
                             torch_v = torch.as_tensor(v, dtype=torch.float32, device=device)
                             if k in cfg.algo.cnn_keys.encoder:
-                                torch_v = torch_v.view(cfg.env.num_envs, -1, *v.shape[-2:])
+                                torch_v = torch_v.view(-1, *v.shape[-2:])
                                 torch_v = torch_v / 255.0 - 0.5
                             real_next_obs[k][i] = torch_v
                     with torch.no_grad():
@@ -480,7 +480,7 @@ def trainer(
 
         # Start training
         with timer(
-            "Time/train_time", SumMetric(sync_on_compute=cfg.metric.sync_on_compute, process_group=optimization_pg)
+            "Time/train_time", SumMetric, sync_on_compute=cfg.metric.sync_on_compute, process_group=optimization_pg
         ):
             # The Join context is needed because there can be the possibility
             # that some ranks receive less data

sheeprl/algos/ppo_recurrent/ppo_recurrent.py

@@ -315,7 +315,7 @@ def main(fabric: Fabric, cfg: Dict[str, Any]):
                         for k, v in info["final_observation"][truncated_env].items():
                             torch_v = torch.as_tensor(v, dtype=torch.float32, device=device)
                             if k in cfg.algo.cnn_keys.encoder:
-                                torch_v = torch_v.view(1, len(truncated_envs), -1, *torch_obs.shape[-2:]) / 255.0 - 0.5
+                                torch_v = torch_v.view(1, 1, -1, *torch_v.shape[-2:]) / 255.0 - 0.5
                             real_next_obs[k][0, i] = torch_v
                     with torch.no_grad():
                         feat = agent.module.feature_extractor(real_next_obs)

sheeprl/algos/sac/sac_decoupled.py

@@ -453,7 +453,7 @@ def trainer(
 
         # Start training
         with timer(
-            "Time/train_time", SumMetric(sync_on_compute=cfg.metric.sync_on_compute, process_group=optimization_pg)
+            "Time/train_time", SumMetric, sync_on_compute=cfg.metric.sync_on_compute, process_group=optimization_pg
         ):
             for batch_idxes in sampler:
                 train(

sheeprl/configs/algo/ppo_recurrent.yaml

@@ -18,11 +18,11 @@ per_rank_sequence_length: ???
 # Model related parameters
 mlp_layers: 1
 layer_norm: True
-dense_units: 128
+dense_units: 256
 dense_act: torch.nn.ReLU
 rnn:
   lstm:
-    hidden_size: 64
+    hidden_size: 128
   pre_rnn_mlp:
     bias: True
     apply: False
@@ -36,7 +36,7 @@ rnn:
     layer_norm: ${algo.layer_norm}
     dense_units: ${algo.rnn.lstm.hidden_size}
 encoder:
-  dense_units: 64
+  dense_units: 128
 
 # Optimizer related parameters
 optimizer:

sheeprl/configs/exp/ppo_recurrent.yaml

@@ -7,17 +7,15 @@ defaults:
   - _self_
 
 algo:
-  per_rank_num_batches: 4
-  per_rank_sequence_length: 8
-  total_steps: 409600
-  rollout_steps: 256
-  update_epochs: 4
+  per_rank_num_batches: 8
+  per_rank_sequence_length: 16
+  total_steps: 409000
+  rollout_steps: 512
+  update_epochs: 8
 
 # Environment
 env:
-  id: CartPole-v1
   num_envs: 16
-  mask_velocities: True
 
 buffer:
   memmap: False

sheeprl/envs/minedojo.py

@@ -192,7 +192,7 @@ def _convert_action(self, action: np.ndarray) -> np.ndarray:
                 self._sticky_attack_counter -= 1
             # it the selected action is not attack, then the agent stops the sticky attack
             elif converted_action[5] != 3:
-                self._sticky_attack = 0
+                self._sticky_attack_counter = 0
         if self._sticky_jump:
             # 2 is the index of the jump/sneak/sprint actions, 1 is the value for the jump action
             if converted_action[2] == 1: