Merge branch 'main' into feature/add-multitask-dit

docs/source/using_dataset_tools.mdx

@@ -163,3 +163,41 @@ lerobot-edit-dataset \
 ```
 
 There is also a tool for adding features to a dataset that is not yet covered in `lerobot-edit-dataset`.
+
+# Dataset Visualization
+
+## Online Visualization
+
+When you record a dataset using `lerobot`, it automatically uploads to the Hugging Face Hub unless you specify otherwise. To view the dataset online, use our **LeRobot Dataset Visualizer**, available at:
+https://huggingface.co/spaces/lerobot/visualize_dataset
+
+## Local Visualization
+
+You can also visualize episodes from a dataset locally using our command-line tool.
+
+**From the Hugging Face Hub:**
+
+```bash
+lerobot-dataset-viz \
+    --repo-id lerobot/pusht \
+    --episode-index 0
+```
+
+**From a local folder:**
+Add the `--root` option and set `--mode local`. For example, to search in `./my_local_data_dir/lerobot/pusht`:
+
+```bash
+lerobot-dataset-viz \
+    --repo-id lerobot/pusht \
+    --root ./my_local_data_dir \
+    --mode local \
+    --episode-index 0
+```
+
+Once executed, the tool opens `rerun.io` and displays the camera streams, robot states, and actions for the selected episode.
+
+For advanced usage—including visualizing datasets stored on a remote server—run:
+
+```bash
+lerobot-dataset-viz --help
+```

src/lerobot/policies/smolvla/modeling_smolvla.py

@@ -527,6 +527,7 @@ class VLAFlowMatching(nn.Module):
             num_vlm_layers=self.config.num_vlm_layers,
             self_attn_every_n_layers=self.config.self_attn_every_n_layers,
             expert_width_multiplier=self.config.expert_width_multiplier,
+            device=self.config.device if self.config.device is not None else "auto",
         )
         self.state_proj = nn.Linear(
             self.config.max_state_dim, self.vlm_with_expert.config.text_config.hidden_size

src/lerobot/scripts/lerobot_info.py

@@ -27,6 +27,25 @@ lerobot-info
 
 import importlib
 import platform
+import shutil
+import subprocess
+from importlib.metadata import PackageNotFoundError, distribution
+
+PACKAGE_NAME = "lerobot"
+
+
+def get_ffmpeg_version() -> str:
+    """Get the ffmpeg version if installed, otherwise return 'N/A'."""
+    command_path = shutil.which("ffmpeg")
+    if command_path is None:
+        return "N/A"
+    try:
+        result = subprocess.run([command_path, "-version"], capture_output=True, text=True, check=True)
+        first_line = result.stdout.splitlines()[0]
+        version_info = first_line.split(" ")[2]
+        return version_info
+    except (subprocess.SubprocessError, IndexError):
+        return "Installed (version parsing failed)"
 
 
 def get_package_version(package_name: str) -> str:
@@ -38,16 +57,17 @@ def get_package_version(package_name: str) -> str:
         return "N/A"
 
 
-def get_sys_info() -> dict:
+def get_sys_info() -> dict[str, str]:
     """Run this to get basic system info to help for tracking issues & bugs."""
     # General package versions
     info = {
-        "lerobot version": get_package_version("lerobot"),
+        "LeRobot version": get_package_version(PACKAGE_NAME),
         "Platform": platform.platform(),
         "Python version": platform.python_version(),
         "Huggingface Hub version": get_package_version("huggingface_hub"),
         "Datasets version": get_package_version("datasets"),
         "Numpy version": get_package_version("numpy"),
+        "FFmpeg version": get_ffmpeg_version(),
     }
 
     # PyTorch and GPU specific information
@@ -58,10 +78,10 @@ def get_sys_info() -> dict:
     try:
         import torch
 
-        torch_version = torch.__version__
+        torch_version = str(torch.__version__)
         torch_cuda_available = torch.cuda.is_available()
         if torch_cuda_available:
-            cuda_version = torch.version.cuda
+            cuda_version = str(torch.version.cuda)
             # Gets the name of the first available GPU
             gpu_model = torch.cuda.get_device_name(0)
     except ImportError:
@@ -71,24 +91,34 @@ def get_sys_info() -> dict:
     info.update(
         {
             "PyTorch version": torch_version,
-            "Is PyTorch built with CUDA support?": torch_cuda_available,
+            "Is PyTorch built with CUDA support?": str(torch_cuda_available),
             "Cuda version": cuda_version,
             "GPU model": gpu_model,
             "Using GPU in script?": "<fill in>",
         }
     )
+    scripts = "N/A"
+    try:
+        dist = distribution(PACKAGE_NAME)
+        scripts = [ep.name for ep in dist.entry_points if ep.group == "console_scripts"]
+    except PackageNotFoundError:
+        pass
+
+    info.update({f"{PACKAGE_NAME} scripts": str(scripts)})
 
     return info
 
 
-def format_dict_for_markdown(d: dict) -> str:
+def format_dict_for_markdown(d: dict[str, str]) -> str:
     """Formats a dictionary into a markdown-friendly bulleted list."""
     return "\n".join([f"- {prop}: {val}" for prop, val in d.items()])
 
 
 def main():
+    """
+    Main function to print system info in markdown format.
+    """
     system_info = get_sys_info()
-    print("\nCopy-and-paste the text below in your GitHub issue and FILL OUT the last point.\n")
     print(format_dict_for_markdown(system_info))