Merge remote-tracking branch 'origin/main' into release/0.3.0

Author: vmoens

.github/unittest/linux_examples/scripts/run_test.sh

@@ -45,6 +45,7 @@ python .github/unittest/helpers/coverage_run_parallel.py examples/decision_trans
   optim.updates_per_episode=3 \
   optim.warmup_steps=10 \
   optim.device=cuda:0 \
+  env.backend=gymnasium \
   logger.backend=
 python .github/unittest/helpers/coverage_run_parallel.py examples/iql/iql_offline.py \
   optim.gradient_steps=55 \

examples/a2c/a2c_atari.py

@@ -93,7 +93,14 @@ def main(cfg: "DictConfig"):  # noqa: F821
     if cfg.logger.backend:
         exp_name = generate_exp_name("A2C", f"{cfg.logger.exp_name}_{cfg.env.env_name}")
         logger = get_logger(
-            cfg.logger.backend, logger_name="a2c", experiment_name=exp_name
+            cfg.logger.backend,
+            logger_name="a2c",
+            experiment_name=exp_name,
+            wandb_kwargs={
+                "config": dict(cfg),
+                "project": cfg.logger.project_name,
+                "group": cfg.logger.group_name,
+            },
         )
 
     # Create test environment

examples/a2c/a2c_mujoco.py

@@ -79,7 +79,14 @@ def main(cfg: "DictConfig"):  # noqa: F821
     if cfg.logger.backend:
         exp_name = generate_exp_name("A2C", f"{cfg.logger.exp_name}_{cfg.env.env_name}")
         logger = get_logger(
-            cfg.logger.backend, logger_name="a2c", experiment_name=exp_name
+            cfg.logger.backend,
+            logger_name="a2c",
+            experiment_name=exp_name,
+            wandb_kwargs={
+                "config": dict(cfg),
+                "project": cfg.logger.project_name,
+                "group": cfg.logger.group_name,
+            },
         )
 
     # Create test environment

examples/a2c/config_atari.yaml

@@ -11,6 +11,8 @@ collector:
 # logger
 logger:
   backend: wandb
+  project_name: torchrl_example_a2c
+  group_name: null
   exp_name: Atari_Schulman17
   test_interval: 40_000_000
   num_test_episodes: 3

examples/a2c/config_mujoco.yaml

@@ -1,6 +1,6 @@
 # task and env
 env:
-  env_name: HalfCheetah-v3
+  env_name: HalfCheetah-v4
 
 # collector
 collector:
@@ -10,6 +10,8 @@ collector:
 # logger
 logger:
   backend: wandb
+  project_name: torchrl_example_a2c
+  group_name: null
   exp_name: Mujoco_Schulman17
   test_interval: 1_000_000
   num_test_episodes: 5

examples/bandits/README.md

@@ -0,0 +1,7 @@
+# Bandits example
+
+## Note:
+This example is not included in the benchmarked results of the current release (v0.3). The intention is to include it in the
+benchmarking of future releases, to ensure that it can be successfully run with the release code and that the
+results are consistent. For now, be aware that this additional check has not been performed in the case of this
+specific example.

examples/cql/cql_offline.py

@@ -32,14 +32,19 @@
 @hydra.main(config_path=".", config_name="offline_config", version_base="1.1")
 def main(cfg: "DictConfig"):  # noqa: F821
     # Create logger
-    exp_name = generate_exp_name("CQL-offline", cfg.env.exp_name)
+    exp_name = generate_exp_name("CQL-offline", cfg.logger.exp_name)
     logger = None
     if cfg.logger.backend:
         logger = get_logger(
             logger_type=cfg.logger.backend,
             logger_name="cql_logging",
             experiment_name=exp_name,
-            wandb_kwargs={"mode": cfg.logger.mode, "config": cfg},
+            wandb_kwargs={
+                "mode": cfg.logger.mode,
+                "config": dict(cfg),
+                "project": cfg.logger.project_name,
+                "group": cfg.logger.group_name,
+            },
         )
     # Set seeds
     torch.manual_seed(cfg.env.seed)

examples/cql/cql_online.py

@@ -36,14 +36,19 @@
 @hydra.main(version_base="1.1", config_path=".", config_name="online_config")
 def main(cfg: "DictConfig"):  # noqa: F821
     # Create logger
-    exp_name = generate_exp_name("CQL-online", cfg.env.exp_name)
+    exp_name = generate_exp_name("CQL-online", cfg.logger.exp_name)
     logger = None
     if cfg.logger.backend:
         logger = get_logger(
             logger_type=cfg.logger.backend,
             logger_name="cql_logging",
             experiment_name=exp_name,
-            wandb_kwargs={"mode": cfg.logger.mode, "config": cfg},
+            wandb_kwargs={
+                "mode": cfg.logger.mode,
+                "config": dict(cfg),
+                "project": cfg.logger.project_name,
+                "group": cfg.logger.group_name,
+            },
         )
 
     # Set seeds

examples/cql/discrete_cql_config.yaml

@@ -3,7 +3,6 @@ env:
   name: CartPole-v1
   task: ""
   backend: gym
-  exp_name: cql_cartpole_gym
   n_samples_stats: 1000
   max_episode_steps: 200
   seed: 0
@@ -24,6 +23,9 @@ collector:
 # Logger
 logger:
   backend: wandb
+  project_name: torchrl_example_cql
+  group_name: null
+  exp_name: cql_cartpole_gym
   log_interval: 5000 # record interval in frames
   eval_steps: 200
   mode: online

examples/cql/discrete_cql_online.py

@@ -38,14 +38,18 @@ def main(cfg: "DictConfig"):  # noqa: F821
     device = torch.device(cfg.optim.device)
 
     # Create logger
-    exp_name = generate_exp_name("DiscreteCQL", cfg.env.exp_name)
+    exp_name = generate_exp_name("DiscreteCQL", cfg.logger.exp_name)
     logger = None
     if cfg.logger.backend:
         logger = get_logger(
             logger_type=cfg.logger.backend,
             logger_name="discretecql_logging",
             experiment_name=exp_name,
-            wandb_kwargs={"mode": cfg.logger.mode, "config": cfg},
+            wandb_kwargs={
+                "mode": cfg.logger.mode,
+                "config": dict(cfg),
+                "project": cfg.logger.project_name,
+            },
         )
 
     # Set seeds