Fixes TiledCamera data types and rlgames training on CPU (#3808)

# Description

We were incorrectly converting all numpy array data in the TiledCamera
class into uint8 type warp arrays when simulation device is set to CPU.
Some annotations like depth are float32 while segmentation data is
uint32. The correct behavior should convert to warp arrays depending on
the input data type of the numpy array.

Additionally, rlgames configs were set to cuda device by default but
were not being overridden when users specify the simulation device to
CPU through cmdline. We should propagate the device setting to the
rlgames configs so that we can run training on the same device, similar
to how RSL RL is set up.

Fixes #3526 

## Type of change

- Bug fix (non-breaking change which fixes an issue)


## Checklist

- [x] I have read and understood the [contribution
guidelines](https://isaac-sim.github.io/IsaacLab/main/source/refs/contributing.html)
- [x] I have run the [`pre-commit` checks](https://pre-commit.com/) with
`./isaaclab.sh --format`
- [x] I have made corresponding changes to the documentation
- [x] My changes generate no new warnings
- [x] I have added tests that prove my fix is effective or that my
feature works
- [x] I have updated the changelog and the corresponding version in the
extension's `config/extension.toml` file
- [x] I have added my name to the `CONTRIBUTORS.md` or my name already
exists there

<!--
As you go through the checklist above, you can mark something as done by
putting an x character in it

For example,
- [x] I have done this task
- [ ] I have not done this task
-->

scripts/benchmarks/benchmark_non_rl.py

@@ -115,6 +115,13 @@ def main(env_cfg: ManagerBasedRLEnvCfg | DirectRLEnvCfg | DirectMARLEnvCfg, agen
     env_cfg.sim.device = args_cli.device if args_cli.device is not None else env_cfg.sim.device
     env_cfg.seed = args_cli.seed
 
+    # check for invalid combination of CPU device with distributed training
+    if args_cli.distributed and args_cli.device is not None and "cpu" in args_cli.device:
+        raise ValueError(
+            "Distributed training is not supported when using CPU device. "
+            "Please use GPU device (e.g., --device cuda) for distributed training."
+        )
+
     # process distributed
     world_size = 1
     world_rank = 0

scripts/benchmarks/benchmark_rlgames.py

@@ -128,6 +128,17 @@ def main(env_cfg: ManagerBasedRLEnvCfg | DirectRLEnvCfg | DirectMARLEnvCfg, agen
     # override configurations with non-hydra CLI arguments
     env_cfg.scene.num_envs = args_cli.num_envs if args_cli.num_envs is not None else env_cfg.scene.num_envs
     env_cfg.sim.device = args_cli.device if args_cli.device is not None else env_cfg.sim.device
+    # check for invalid combination of CPU device with distributed training
+    if args_cli.distributed and args_cli.device is not None and "cpu" in args_cli.device:
+        raise ValueError(
+            "Distributed training is not supported when using CPU device. "
+            "Please use GPU device (e.g., --device cuda) for distributed training."
+        )
+
+    # update agent device to match simulation device
+    if args_cli.device is not None:
+        agent_cfg["params"]["config"]["device"] = args_cli.device
+        agent_cfg["params"]["config"]["device_name"] = args_cli.device
 
     # randomly sample a seed if seed = -1
     if args_cli.seed == -1:

scripts/benchmarks/benchmark_rsl_rl.py

@@ -140,6 +140,12 @@ def main(env_cfg: ManagerBasedRLEnvCfg | DirectRLEnvCfg | DirectMARLEnvCfg, agen
     # note: certain randomizations occur in the environment initialization so we set the seed here
     env_cfg.seed = agent_cfg.seed
     env_cfg.sim.device = args_cli.device if args_cli.device is not None else env_cfg.sim.device
+    # check for invalid combination of CPU device with distributed training
+    if args_cli.distributed and args_cli.device is not None and "cpu" in args_cli.device:
+        raise ValueError(
+            "Distributed training is not supported when using CPU device. "
+            "Please use GPU device (e.g., --device cuda) for distributed training."
+        )
 
     # multi-gpu training configuration
     world_rank = 0

scripts/reinforcement_learning/rl_games/play.py

@@ -95,6 +95,10 @@ def main(env_cfg: ManagerBasedRLEnvCfg | DirectRLEnvCfg | DirectMARLEnvCfg, agen
     # override configurations with non-hydra CLI arguments
     env_cfg.scene.num_envs = args_cli.num_envs if args_cli.num_envs is not None else env_cfg.scene.num_envs
     env_cfg.sim.device = args_cli.device if args_cli.device is not None else env_cfg.sim.device
+    # update agent device to match simulation device
+    if args_cli.device is not None:
+        agent_cfg["params"]["config"]["device"] = args_cli.device
+        agent_cfg["params"]["config"]["device_name"] = args_cli.device
 
     # randomly sample a seed if seed = -1
     if args_cli.seed == -1:

scripts/reinforcement_learning/rl_games/train.py

@@ -95,6 +95,17 @@ def main(env_cfg: ManagerBasedRLEnvCfg | DirectRLEnvCfg | DirectMARLEnvCfg, agen
     # override configurations with non-hydra CLI arguments
     env_cfg.scene.num_envs = args_cli.num_envs if args_cli.num_envs is not None else env_cfg.scene.num_envs
     env_cfg.sim.device = args_cli.device if args_cli.device is not None else env_cfg.sim.device
+    # check for invalid combination of CPU device with distributed training
+    if args_cli.distributed and args_cli.device is not None and "cpu" in args_cli.device:
+        raise ValueError(
+            "Distributed training is not supported when using CPU device. "
+            "Please use GPU device (e.g., --device cuda) for distributed training."
+        )
+
+    # update agent device to match simulation device
+    if args_cli.device is not None:
+        agent_cfg["params"]["config"]["device"] = args_cli.device
+        agent_cfg["params"]["config"]["device_name"] = args_cli.device
 
     # randomly sample a seed if seed = -1
     if args_cli.seed == -1:

scripts/reinforcement_learning/rsl_rl/train.py

@@ -118,6 +118,12 @@ def main(env_cfg: ManagerBasedRLEnvCfg | DirectRLEnvCfg | DirectMARLEnvCfg, agen
     # note: certain randomizations occur in the environment initialization so we set the seed here
     env_cfg.seed = agent_cfg.seed
     env_cfg.sim.device = args_cli.device if args_cli.device is not None else env_cfg.sim.device
+    # check for invalid combination of CPU device with distributed training
+    if args_cli.distributed and args_cli.device is not None and "cpu" in args_cli.device:
+        raise ValueError(
+            "Distributed training is not supported when using CPU device. "
+            "Please use GPU device (e.g., --device cuda) for distributed training."
+        )
 
     # multi-gpu training configuration
     if args_cli.distributed:

scripts/reinforcement_learning/skrl/train.py

@@ -129,6 +129,13 @@ def main(env_cfg: ManagerBasedRLEnvCfg | DirectRLEnvCfg | DirectMARLEnvCfg, agen
     env_cfg.scene.num_envs = args_cli.num_envs if args_cli.num_envs is not None else env_cfg.scene.num_envs
     env_cfg.sim.device = args_cli.device if args_cli.device is not None else env_cfg.sim.device
 
+    # check for invalid combination of CPU device with distributed training
+    if args_cli.distributed and args_cli.device is not None and "cpu" in args_cli.device:
+        raise ValueError(
+            "Distributed training is not supported when using CPU device. "
+            "Please use GPU device (e.g., --device cuda) for distributed training."
+        )
+
     # multi-gpu training config
     if args_cli.distributed:
         env_cfg.sim.device = f"cuda:{app_launcher.local_rank}"

source/isaaclab/config/extension.toml

 [package]
 
 # Note: Semantic Versioning is used: https://semver.org/
-version = "0.47.1"
+version = "0.47.2"
 
 # Description
 title = "Isaac Lab framework for Robot Learning"

source/isaaclab/docs/CHANGELOG.rst

 Changelog
 ---------
 
+0.47.2 (2025-10-22)
+~~~~~~~~~~~~~~~~~~~
+
+Changed
+^^^^^^^
+
+* Fixed the data type conversion in :class:`~isaaclab.sensors.tiled_camera.TiledCamera` to
+  support the correct data type when converting from numpy arrays to warp arrays on the CPU.
+
+
 0.47.1 (2025-10-17)
 ~~~~~~~~~~~~~~~~~~~
 

source/isaaclab/isaaclab/sensors/camera/tiled_camera.py

@@ -248,7 +248,9 @@ class TiledCamera(Camera):
 
             # convert data buffer to warp array
             if isinstance(tiled_data_buffer, np.ndarray):
-                tiled_data_buffer = wp.array(tiled_data_buffer, device=self.device, dtype=wp.uint8)
+                # Let warp infer the dtype from numpy array instead of hardcoding uint8
+                # Different annotators return different dtypes: RGB(uint8), depth(float32), segmentation(uint32)
+                tiled_data_buffer = wp.array(tiled_data_buffer, device=self.device)
             else:
                 tiled_data_buffer = tiled_data_buffer.to(device=self.device)