amend

pytorch · vmoens · Jan 31, 2024 · Jan 19, 2024 · Jan 31, 2024 · Jan 31, 2024
commit 62b1dc8434da524855c9736c015d091d4c79c3a7
diff --git a/examples/a2c/a2c_atari.py b/examples/a2c/a2c_atari.py
@@ -31,7 +31,7 @@ def main(cfg: "DictConfig"):  # noqa: F821
     total_frames = cfg.collector.total_frames // frame_skip
     frames_per_batch = cfg.collector.frames_per_batch // frame_skip
     mini_batch_size = cfg.loss.mini_batch_size // frame_skip
-    test_interval = cfg.torchrl_logger.test_interval // frame_skip
+    test_interval = cfg.logger.test_interval // frame_skip
 
     # Create models (check utils_atari.py)
     actor, critic, critic_head = make_ppo_models(cfg.env.env_name)
@@ -87,20 +87,20 @@ def main(cfg: "DictConfig"):  # noqa: F821
         eps=cfg.optim.eps,
     )
 
-    # Create torchrl_logger
+    # Create logger
     logger = None
-    if cfg.torchrl_logger.backend:
+    if cfg.logger.backend:
         exp_name = generate_exp_name(
-            "A2C", f"{cfg.torchrl_logger.exp_name}_{cfg.env.env_name}"
+            "A2C", f"{cfg.logger.exp_name}_{cfg.env.env_name}"
         )
         logger = get_logger(
-            cfg.torchrl_logger.backend,
+            cfg.logger.backend,
             logger_name="a2c",
             experiment_name=exp_name,
             wandb_kwargs={
                 "config": dict(cfg),
-                "project": cfg.torchrl_logger.project_name,
-                "group": cfg.torchrl_logger.group_name,
+                "project": cfg.logger.project_name,
+                "group": cfg.logger.group_name,
             },
         )
 
@@ -201,7 +201,7 @@ def main(cfg: "DictConfig"):  # noqa: F821
                 actor.eval()
                 eval_start = time.time()
                 test_rewards = eval_model(
-                    actor, test_env, num_episodes=cfg.torchrl_logger.num_test_episodes
+                    actor, test_env, num_episodes=cfg.logger.num_test_episodes
                 )
                 eval_time = time.time() - eval_start
                 log_info.update(

diff --git a/examples/a2c/a2c_mujoco.py b/examples/a2c/a2c_mujoco.py
@@ -73,20 +73,20 @@ def main(cfg: "DictConfig"):  # noqa: F821
     actor_optim = torch.optim.Adam(actor.parameters(), lr=cfg.optim.lr)
     critic_optim = torch.optim.Adam(critic.parameters(), lr=cfg.optim.lr)
 
-    # Create torchrl_logger
+    # Create logger
     logger = None
-    if cfg.torchrl_logger.backend:
+    if cfg.logger.backend:
         exp_name = generate_exp_name(
-            "A2C", f"{cfg.torchrl_logger.exp_name}_{cfg.env.env_name}"
+            "A2C", f"{cfg.logger.exp_name}_{cfg.env.env_name}"
         )
         logger = get_logger(
-            cfg.torchrl_logger.backend,
+            cfg.logger.backend,
             logger_name="a2c",
             experiment_name=exp_name,
             wandb_kwargs={
                 "config": dict(cfg),
-                "project": cfg.torchrl_logger.project_name,
-                "group": cfg.torchrl_logger.group_name,
+                "project": cfg.logger.project_name,
+                "group": cfg.logger.group_name,
             },
         )
 
@@ -180,13 +180,13 @@ def main(cfg: "DictConfig"):  # noqa: F821
 
         # Get test rewards
         with torch.no_grad(), set_exploration_type(ExplorationType.MODE):
-            if ((i - 1) * frames_in_batch) // cfg.torchrl_logger.test_interval < (
+            if ((i - 1) * frames_in_batch) // cfg.logger.test_interval < (
                 i * frames_in_batch
-            ) // cfg.torchrl_logger.test_interval:
+            ) // cfg.logger.test_interval:
                 actor.eval()
                 eval_start = time.time()
                 test_rewards = eval_model(
-                    actor, test_env, num_episodes=cfg.torchrl_logger.num_test_episodes
+                    actor, test_env, num_episodes=cfg.logger.num_test_episodes
                 )
                 eval_time = time.time() - eval_start
                 log_info.update(

diff --git a/examples/a2c/config_atari.yaml b/examples/a2c/config_atari.yaml
@@ -8,7 +8,7 @@ collector:
   frames_per_batch: 80
   total_frames: 40_000_000
 
-# torchrl_logger
+# logger
 logger:
   backend: wandb
   project_name: torchrl_example_a2c

diff --git a/examples/a2c/config_mujoco.yaml b/examples/a2c/config_mujoco.yaml
@@ -7,7 +7,7 @@ collector:
   frames_per_batch: 64
   total_frames: 1_000_000
 
-# torchrl_logger
+# logger
 logger:
   backend: wandb
   project_name: torchrl_example_a2c

diff --git a/examples/cql/cql_offline.py b/examples/cql/cql_offline.py
@@ -31,19 +31,19 @@
 
 @hydra.main(config_path=".", config_name="offline_config", version_base="1.1")
 def main(cfg: "DictConfig"):  # noqa: F821
-    # Create torchrl_logger
-    exp_name = generate_exp_name("CQL-offline", cfg.torchrl_logger.exp_name)
+    # Create logger
+    exp_name = generate_exp_name("CQL-offline", cfg.logger.exp_name)
     logger = None
-    if cfg.torchrl_logger.backend:
+    if cfg.logger.backend:
         logger = get_logger(
-            logger_type=cfg.torchrl_logger.backend,
+            logger_type=cfg.logger.backend,
             logger_name="cql_logging",
             experiment_name=exp_name,
             wandb_kwargs={
-                "mode": cfg.torchrl_logger.mode,
+                "mode": cfg.logger.mode,
                 "config": dict(cfg),
-                "project": cfg.torchrl_logger.project_name,
-                "group": cfg.torchrl_logger.group_name,
+                "project": cfg.logger.project_name,
+                "group": cfg.logger.group_name,
             },
         )
     # Set seeds
@@ -52,7 +52,7 @@ def main(cfg: "DictConfig"):  # noqa: F821
     device = torch.device(cfg.optim.device)
 
     # Create env
-    train_env, eval_env = make_environment(cfg, cfg.torchrl_logger.eval_envs)
+    train_env, eval_env = make_environment(cfg, cfg.logger.eval_envs)
 
     # Create replay buffer
     replay_buffer = make_offline_replay_buffer(cfg.replay_buffer)
@@ -75,8 +75,8 @@ def main(cfg: "DictConfig"):  # noqa: F821
 
     gradient_steps = cfg.optim.gradient_steps
     policy_eval_start = cfg.optim.policy_eval_start
-    evaluation_interval = cfg.torchrl_logger.eval_iter
-    eval_steps = cfg.torchrl_logger.eval_steps
+    evaluation_interval = cfg.logger.eval_iter
+    eval_steps = cfg.logger.eval_steps
 
     # Training loop
     start_time = time.time()

diff --git a/examples/cql/cql_online.py b/examples/cql/cql_online.py
@@ -35,19 +35,19 @@
 
 @hydra.main(version_base="1.1", config_path=".", config_name="online_config")
 def main(cfg: "DictConfig"):  # noqa: F821
-    # Create torchrl_logger
-    exp_name = generate_exp_name("CQL-online", cfg.torchrl_logger.exp_name)
+    # Create logger
+    exp_name = generate_exp_name("CQL-online", cfg.logger.exp_name)
     logger = None
-    if cfg.torchrl_logger.backend:
+    if cfg.logger.backend:
         logger = get_logger(
-            logger_type=cfg.torchrl_logger.backend,
+            logger_type=cfg.logger.backend,
             logger_name="cql_logging",
             experiment_name=exp_name,
             wandb_kwargs={
-                "mode": cfg.torchrl_logger.mode,
+                "mode": cfg.logger.mode,
                 "config": dict(cfg),
-                "project": cfg.torchrl_logger.project_name,
-                "group": cfg.torchrl_logger.group_name,
+                "project": cfg.logger.project_name,
+                "group": cfg.logger.group_name,
             },
         )
 
@@ -99,7 +99,7 @@ def main(cfg: "DictConfig"):  # noqa: F821
         * cfg.optim.utd_ratio
     )
     prb = cfg.replay_buffer.prb
-    eval_iter = cfg.torchrl_logger.eval_iter
+    eval_iter = cfg.logger.eval_iter
     frames_per_batch = cfg.collector.frames_per_batch
     eval_rollout_steps = cfg.collector.max_frames_per_traj
 

diff --git a/examples/cql/discrete_cql_online.py b/examples/cql/discrete_cql_online.py
@@ -37,18 +37,18 @@
 def main(cfg: "DictConfig"):  # noqa: F821
     device = torch.device(cfg.optim.device)
 
-    # Create torchrl_logger
-    exp_name = generate_exp_name("DiscreteCQL", cfg.torchrl_logger.exp_name)
+    # Create logger
+    exp_name = generate_exp_name("DiscreteCQL", cfg.logger.exp_name)
     logger = None
-    if cfg.torchrl_logger.backend:
+    if cfg.logger.backend:
         logger = get_logger(
-            logger_type=cfg.torchrl_logger.backend,
+            logger_type=cfg.logger.backend,
             logger_name="discretecql_logging",
             experiment_name=exp_name,
             wandb_kwargs={
-                "mode": cfg.torchrl_logger.mode,
+                "mode": cfg.logger.mode,
                 "config": dict(cfg),
-                "project": cfg.torchrl_logger.project_name,
+                "project": cfg.logger.project_name,
             },
         )
 
@@ -92,7 +92,7 @@ def main(cfg: "DictConfig"):  # noqa: F821
     )
     prb = cfg.replay_buffer.prb
     eval_rollout_steps = cfg.env.max_episode_steps
-    eval_iter = cfg.torchrl_logger.eval_iter
+    eval_iter = cfg.logger.eval_iter
     frames_per_batch = cfg.collector.frames_per_batch
 
     start_time = sampling_start = time.time()

diff --git a/examples/cql/offline_config.yaml b/examples/cql/offline_config.yaml
@@ -7,7 +7,7 @@ env:
   seed: 0
   backend: gymnasium
 
-# torchrl_logger
+# logger
 logger:
   backend: wandb
   project_name: torchrl_example_cql

diff --git a/examples/cql/online_config.yaml b/examples/cql/online_config.yaml
@@ -19,7 +19,7 @@ collector:
   max_frames_per_traj: 1000
 
 
-# torchrl_logger
+# logger
 logger:
   backend: wandb
   project_name: torchrl_example_cql

diff --git a/examples/ddpg/ddpg.py b/examples/ddpg/ddpg.py
@@ -37,19 +37,19 @@
 def main(cfg: "DictConfig"):  # noqa: F821
     device = torch.device(cfg.network.device)
 
-    # Create torchrl_logger
-    exp_name = generate_exp_name("DDPG", cfg.torchrl_logger.exp_name)
+    # Create logger
+    exp_name = generate_exp_name("DDPG", cfg.logger.exp_name)
     logger = None
-    if cfg.torchrl_logger.backend:
+    if cfg.logger.backend:
         logger = get_logger(
-            logger_type=cfg.torchrl_logger.backend,
+            logger_type=cfg.logger.backend,
             logger_name="ddpg_logging",
             experiment_name=exp_name,
             wandb_kwargs={
-                "mode": cfg.torchrl_logger.mode,
+                "mode": cfg.logger.mode,
                 "config": dict(cfg),
-                "project": cfg.torchrl_logger.project_name,
-                "group": cfg.torchrl_logger.group_name,
+                "project": cfg.logger.project_name,
+                "group": cfg.logger.group_name,
             },
         )
 
@@ -94,7 +94,7 @@ def main(cfg: "DictConfig"):  # noqa: F821
     )
     prb = cfg.replay_buffer.prb
     frames_per_batch = cfg.collector.frames_per_batch
-    eval_iter = cfg.torchrl_logger.eval_iter
+    eval_iter = cfg.logger.eval_iter
     eval_rollout_steps = cfg.env.max_episode_steps
 
     sampling_start = time.time()

diff --git a/examples/decision_transformer/dt.py b/examples/decision_transformer/dt.py
@@ -75,8 +75,8 @@ def main(cfg: "DictConfig"):  # noqa: F821
 
     pretrain_gradient_steps = cfg.optim.pretrain_gradient_steps
     clip_grad = cfg.optim.clip_grad
-    eval_steps = cfg.torchrl_logger.eval_steps
-    pretrain_log_interval = cfg.torchrl_logger.pretrain_log_interval
+    eval_steps = cfg.logger.eval_steps
+    pretrain_log_interval = cfg.logger.pretrain_log_interval
     reward_scaling = cfg.env.reward_scaling
 
     torchrl_logger.info(" ***Pretraining*** ")

diff --git a/examples/decision_transformer/online_dt.py b/examples/decision_transformer/online_dt.py
@@ -77,8 +77,8 @@ def main(cfg: "DictConfig"):  # noqa: F821
 
     pretrain_gradient_steps = cfg.optim.pretrain_gradient_steps
     clip_grad = cfg.optim.clip_grad
-    eval_steps = cfg.torchrl_logger.eval_steps
-    pretrain_log_interval = cfg.torchrl_logger.pretrain_log_interval
+    eval_steps = cfg.logger.eval_steps
+    pretrain_log_interval = cfg.logger.pretrain_log_interval
     reward_scaling = cfg.env.reward_scaling
 
     torchrl_logger.info(" ***Pretraining*** ")

diff --git a/examples/decision_transformer/utils.py b/examples/decision_transformer/utils.py
@@ -493,19 +493,19 @@ def make_dt_optimizer(optim_cfg, loss_module):
 
 
 def make_logger(cfg):
-    if not cfg.torchrl_logger.backend:
+    if not cfg.logger.backend:
         return None
     exp_name = generate_exp_name(
-        cfg.torchrl_logger.model_name, cfg.torchrl_logger.exp_name
+        cfg.logger.model_name, cfg.logger.exp_name
     )
     logger = get_logger(
-        cfg.torchrl_logger.backend,
-        logger_name=cfg.torchrl_logger.model_name,
+        cfg.logger.backend,
+        logger_name=cfg.logger.model_name,
         experiment_name=exp_name,
         wandb_kwargs={
             "config": dict(cfg),
-            "project": cfg.torchrl_logger.project_name,
-            "group": cfg.torchrl_logger.group_name,
+            "project": cfg.logger.project_name,
+            "group": cfg.logger.group_name,
         },
     )
     return logger

diff --git a/examples/discrete_sac/discrete_sac.py b/examples/discrete_sac/discrete_sac.py
@@ -38,18 +38,18 @@ def main(cfg: "DictConfig"):  # noqa: F821
     device = torch.device(cfg.network.device)
 
     # Create torchrl_logger
-    exp_name = generate_exp_name("DiscreteSAC", cfg.torchrl_logger.exp_name)
+    exp_name = generate_exp_name("DiscreteSAC", cfg.logger.exp_name)
     logger = None
-    if cfg.torchrl_logger.backend:
+    if cfg.logger.backend:
         logger = get_logger(
-            logger_type=cfg.torchrl_logger.backend,
+            logger_type=cfg.logger.backend,
             logger_name="DiscreteSAC_logging",
             experiment_name=exp_name,
             wandb_kwargs={
-                "mode": cfg.torchrl_logger.mode,
+                "mode": cfg.logger.mode,
                 "config": dict(cfg),
-                "project": cfg.torchrl_logger.project_name,
-                "group": cfg.torchrl_logger.group_name,
+                "project": cfg.logger.project_name,
+                "group": cfg.logger.group_name,
             },
         )
 
@@ -96,7 +96,7 @@ def main(cfg: "DictConfig"):  # noqa: F821
     )
     prb = cfg.replay_buffer.prb
     eval_rollout_steps = cfg.env.max_episode_steps
-    eval_iter = cfg.torchrl_logger.eval_iter
+    eval_iter = cfg.logger.eval_iter
     frames_per_batch = cfg.collector.frames_per_batch
 
     sampling_start = time.time()

diff --git a/examples/dqn/config_atari.yaml b/examples/dqn/config_atari.yaml
@@ -19,7 +19,7 @@ buffer:
   batch_size: 32
   scratch_dir: null
 
-# torchrl_logger
+# logger
 logger:
   backend: wandb
   project_name: torchrl_example_dqn

diff --git a/examples/dqn/config_cartpole.yaml b/examples/dqn/config_cartpole.yaml
@@ -18,7 +18,7 @@ buffer:
   buffer_size: 10_000
   batch_size: 128
 
-# torchrl_logger
+# logger
 logger:
   backend: wandb
   project_name: torchrl_example_dqn