shared_with_container and files inside have user permissions instead of root

README.md

@@ -65,7 +65,7 @@ There are two main ways to execute configure the conversion process:
 
 1. **NN Archive**:
    Alternatively, you can use an [NN Archive](https://docs.luxonis.com/software-v3/ai-inference/nn-archive/#NN%20Archive) as input. An NN Archive includes a model in one of the supported formats—ONNX (.onnx), OpenVINO IR (.xml and .bin), or TensorFlow Lite (.tflite)—alongside a `config.json` file. The config.json file follows a specific configuration format as described under the `Configuration` section.
-1. **YAML Configuration File (Legacy)**:
+2. **YAML Configuration File (Legacy)**:
    An alternative way to configure the conversion is through a YAML configuration file. For reference, you can check [defaults.yaml](shared_with_container/configs/defaults.yaml) and other examples located in the [shared_with_container/configs](shared_with_container/configs) directory.
 
 **Modifying Settings with Command-Line Arguments**:
@@ -324,6 +324,9 @@ shared_with_container/
 │ ├── resnet18.yaml
 │ └── <configs will be downloaded here>
 │
+├── misc/
+│ └── <internal scratch files>
+│
 ├── models/
 │ ├── resnet18.onnx
 │ └── <models will be downloaded here>
@@ -345,7 +348,9 @@ The converter first searches for files exactly at the provided path. If not foun
 The `output_dir` can be specified using the `--output-dir` CLI argument. If such a directory already exists, the `output_dir_name` will be appended with the current date and time. If not specified, the `output_dir_name` will be autogenerated in the following format: `<model_name>_to_<target>_<date>_<time>`.
 
 > [!NOTE]
-> When running the CLI, `shared_with_container` must be present in the current working directory, and all paths provided to the CLI must be specified relative to the `shared_with_container` directory.
+> When running the `modelconverter` CLI, `shared_with_container` and its standard subdirectories are created automatically in the current working directory if they do not exist yet.
+> With the default CLI, local input files should be placed under `shared_with_container/`, for example `configs/resnet18.yaml` or `models/resnet18.onnx`.
+> Files outside of `shared_with_container/` are not visible to the container unless you mount them separately. Remote paths such as `s3://...` and `gs://...` work as usual.
 
 ### Running ModelConverter
 
@@ -359,11 +364,17 @@ You can run the built image either manually using the `docker run` command or us
    export AWS_S3_ENDPOINT_URL=<your_aws_s3_endpoint_url>
    ```
 
-1. If `shared_with_container` directory doesn't exist on your host, create it.
+2. If using the `modelconverter` CLI, `shared_with_container` is created automatically in the current working directory if it does not exist yet.
+
+- If using `docker run` manually, create `shared_with_container` on the host first so it can be mounted into the container:
+
+  ```bash
+  mkdir -p shared_with_container
+  ```
 
-1. Without remote files, place the model, config, and calibration data in the respective directories (refer [Sharing Files](#sharing-files)).
+3. Without remote files, place the model, config, and calibration data somewhere under `shared_with_container/` (refer [Sharing Files](#sharing-files)).
 
-1. Execute the conversion:
+4. Execute the conversion:
 
 - If using the `modelconverter` CLI:
 
@@ -641,8 +652,8 @@ optionally saves the results to a `.csv` file.
 ModelConverter offers additional analysis tools for the RVC4 platform. The tools provide an in-depth look at the following:
 
 1. The outputs of all layers in comparison to the ground truth ONNX model,
-1. The cycle usage of each layer on an RVC4 device.
-1. Visualizations for fast and easy comparison of multiple models.
+2. The cycle usage of each layer on an RVC4 device.
+3. Visualizations for fast and easy comparison of multiple models.
 
 This gives the user better insight into the successful quantization of a model, helps discover potential speed bottleneck layers, and allows for the comparison of different quantization parameters.
 

modelconverter/__main__.py

@@ -623,6 +623,7 @@ def launcher(
     if in_docker():
         return command(*bound.args, **bound.kwargs)
 
+    init_dirs()
     tag = "dev" if dev else "latest"
 
     target = bound.arguments["target"]

modelconverter/cli/utils.py

@@ -24,6 +24,9 @@
     MISC_DIR,
     MODELS_DIR,
     OUTPUTS_DIR,
+    RUNTIME_CACHE_SUFFIX,
+    RUNTIME_HOME_SUFFIX,
+    SHARED_DIR,
 )
 from modelconverter.utils.hub_requests import Request
 from modelconverter.utils.types import DataType, Encoding, Target
@@ -66,7 +69,18 @@ def get_output_dir_name(
 
 
 def init_dirs() -> None:
-    for p in [CONFIGS_DIR, MODELS_DIR, OUTPUTS_DIR, CALIBRATION_DIR]:
+    for p in [
+        SHARED_DIR,
+        MISC_DIR,
+        # Some tools run inside the container as the host UID and still
+        # expect writable HOME/XDG cache locations.
+        MISC_DIR / RUNTIME_HOME_SUFFIX,
+        MISC_DIR / RUNTIME_CACHE_SUFFIX,
+        CONFIGS_DIR,
+        MODELS_DIR,
+        OUTPUTS_DIR,
+        CALIBRATION_DIR,
+    ]:
         logger.debug(f"Creating {p}")
         p.mkdir(parents=True, exist_ok=True)
 

modelconverter/utils/constants.py

@@ -9,12 +9,16 @@
 OUTPUTS_DIR: Final[Path] = SHARED_DIR / "outputs"
 CALIBRATION_DIR: Final[Path] = SHARED_DIR / "calibration_data"
 MODELS_DIR: Final[Path] = SHARED_DIR / "models"
+RUNTIME_HOME_SUFFIX: Final[str] = "runtime-home"
+RUNTIME_CACHE_SUFFIX: Final[str] = "runtime-cache"
 LOADERS = Registry(name="loaders")
 
 __all__ = [
     "CALIBRATION_DIR",
     "CONFIGS_DIR",
     "MODELS_DIR",
     "OUTPUTS_DIR",
+    "RUNTIME_CACHE_SUFFIX",
+    "RUNTIME_HOME_SUFFIX",
     "SHARED_DIR",
 ]

modelconverter/utils/docker_utils.py

@@ -22,6 +22,14 @@
 
 import docker
 from modelconverter import __version__
+from modelconverter.utils.constants import (
+    RUNTIME_CACHE_SUFFIX,
+    RUNTIME_HOME_SUFFIX,
+)
+
+_CONTAINER_MISC_DIR = "/app/shared_with_container/misc"
+_CONTAINER_USER_HOME = f"{_CONTAINER_MISC_DIR}/{RUNTIME_HOME_SUFFIX}"
+_CONTAINER_USER_CACHE = f"{_CONTAINER_MISC_DIR}/{RUNTIME_CACHE_SUFFIX}"
 
 
 def get_docker_client_from_active_context() -> docker.DockerClient:
@@ -73,6 +81,7 @@
     memory: str | None = None,
     cpus: float | None = None,
 ) -> str:
+    user_spec = _get_current_user_spec()
     config = {
         "services": {
             "modelconverter": {
@@ -104,6 +113,14 @@
             }
         },
     }
+    if user_spec is not None:
+        config["services"]["modelconverter"]["user"] = user_spec
+        config["services"]["modelconverter"]["environment"].update(
+            {
+                "HOME": _CONTAINER_USER_HOME,
+                "XDG_CACHE_HOME": _CONTAINER_USER_CACHE,
+            }
+        )
     limits = {}
     if memory is not None:
         limits["memory"] = memory
@@ -121,6 +138,14 @@
     return yaml.dump(config)
 
 
+def _get_current_user_spec() -> str | None:
+    getuid = getattr(os, "getuid", None)
+    getgid = getattr(os, "getgid", None)
+    if not callable(getuid) or not callable(getgid):
+        return None
+    return f"{getuid()}:{getgid()}"
+
+
 def in_docker() -> bool:
     return "IN_DOCKER" in os.environ
 
@@ -261,7 +286,7 @@
     tmp_path: Path | None = None
     try:
         request = Request(url, headers={"User-Agent": "modelconverter"})  # noqa: S310
         with urlopen(request, timeout=30) as response:  # noqa: S310
             if getattr(response, "status", 200) >= 400:
                 raise RuntimeError(
                     f"HTTP {response.status} while downloading {url}"