Merge branch 'Project-MONAI:dev' into dev

docs/source/mb_specification.rst

@@ -63,12 +63,12 @@ This file contains the metadata information relating to the model, including wha
 * **monai_version**: version of MONAI the bundle was generated on, later versions expected to work.
 * **pytorch_version**: version of Pytorch the bundle was generated on, later versions expected to work.
 * **numpy_version**: version of Numpy the bundle was generated on, later versions expected to work.
-* **optional_packages_version**: dictionary relating optional package names to their versions, these packages are not needed but are recommended to be installed with this stated minimum version.
+* **required_packages_version**: dictionary relating required package names to their versions. These are packages in addition to the base requirements of MONAI which this bundle absolutely needs. For example, if the bundle must load Nifti files the Nibabel package will be required.
 * **task**: plain-language description of what the model is meant to do.
 * **description**: longer form plain-language description of what the model is, what it does, etc.
 * **authors**: state author(s) of the model.
 * **copyright**: state model copyright.
-* **network_data_format**: defines the format, shape, and meaning of inputs and outputs to the model, contains keys "inputs" and "outputs" relating named inputs/outputs to their format specifiers (defined below).
+* **network_data_format**: defines the format, shape, and meaning of inputs and outputs to the (primary) model, contains keys "inputs" and "outputs" relating named inputs/outputs to their format specifiers (defined below). There is also an optional "post_processed_outputs" key stating the format of "outputs" after postprocessing transforms are applied, this is used to describe the final output from the bundle if it varies from the raw network output. These keys can also relate to primitive values (number, string, boolean), instead of the tensor format specified below.
 
 Tensor format specifiers are used to define input and output tensors and their meanings, and must be a dictionary containing at least these keys:
 
@@ -89,6 +89,8 @@ Optional keys:
 * **data_source**: description of where training/validation can be sourced.
 * **data_type**: type of source data used for training/validation.
 * **references**: list of published referenced relating to the model.
+* **supported_apps**: list of supported applications which use bundles, eg. 'monai-label' would be present if the bundle is compatible with MONAI Label applications.
+* **\*_data_format**: defines the format, shape, and meaning of inputs and outputs to additional models which are secondary to the main model. This contains the same sort of information as **network_data_format** which describes networks providing secondary functionality, eg. a localisation network used to identify ROI in an image for cropping before data is sent to the primary network of this bundle.
 
 The format for tensors used as inputs and outputs can be used to specify semantic meaning of these values, and later is used by software handling bundles to determine how to process and interpret this data. There are various types of image data that MONAI is uses, and other data types such as point clouds, dictionary sequences, time signals, and others. The following list is provided as a set of supported definitions of what a tensor "format" is but is not exhaustive and users can provide their own which would be left up to the model users to interpret:
 
@@ -124,7 +126,7 @@ An example JSON metadata file:
       "monai_version": "0.9.0",
       "pytorch_version": "1.10.0",
       "numpy_version": "1.21.2",
-      "optional_packages_version": {"nibabel": "3.2.1"},
+      "required_packages_version": {"nibabel": "3.2.1"},
       "task": "Decathlon spleen segmentation",
       "description": "A pre-trained model for volumetric (3D) segmentation of the spleen from CT image",
       "authors": "MONAI team",

monai/auto3dseg/analyzer.py

@@ -470,7 +470,7 @@ def __call__(self, data: Mapping[Hashable, MetaTensor]) -> dict[Hashable, MetaTe
 
         unique_label = unique(ndas_label)
         if isinstance(ndas_label, (MetaTensor, torch.Tensor)):
-            unique_label = unique_label.data.cpu().numpy()
+            unique_label = unique_label.data.cpu().numpy()  # type: ignore[assignment]
 
         unique_label = unique_label.astype(np.int16).tolist()
 

monai/bundle/scripts.py

@@ -27,7 +27,7 @@
 import torch
 from torch.cuda import is_available
 
-from monai.apps.mmars.mmars import _get_all_ngc_models
+from monai._version import get_versions
 from monai.apps.utils import _basename, download_url, extractall, get_logger
 from monai.bundle.config_item import ConfigComponent
 from monai.bundle.config_parser import ConfigParser
@@ -67,6 +67,9 @@
 DEFAULT_DOWNLOAD_SOURCE = os.environ.get("BUNDLE_DOWNLOAD_SRC", "monaihosting")
 PPRINT_CONFIG_N = 5
 
+MONAI_HOSTING_BASE_URL = "https://api.ngc.nvidia.com/v2/models/nvidia/monaihosting"
+NGC_BASE_URL = "https://api.ngc.nvidia.com/v2/models/nvidia/monaitoolkit"
+
 
 def update_kwargs(args: str | dict | None = None, ignore_none: bool = True, **kwargs: Any) -> dict:
     """
@@ -169,16 +172,19 @@ def _get_git_release_url(repo_owner: str, repo_name: str, tag_name: str, filenam
 
 
 def _get_ngc_bundle_url(model_name: str, version: str) -> str:
-    return f"https://api.ngc.nvidia.com/v2/models/nvidia/monaitoolkit/{model_name.lower()}/versions/{version}/zip"
+    return f"{NGC_BASE_URL}/{model_name.lower()}/versions/{version}/zip"
+
+
+def _get_ngc_private_base_url(repo: str) -> str:
+    return f"https://api.ngc.nvidia.com/v2/{repo}/models"
 
 
 def _get_ngc_private_bundle_url(model_name: str, version: str, repo: str) -> str:
-    return f"https://api.ngc.nvidia.com/v2/{repo}/models/{model_name.lower()}/versions/{version}/zip"
+    return f"{_get_ngc_private_base_url(repo)}/{model_name.lower()}/versions/{version}/zip"
 
 
 def _get_monaihosting_bundle_url(model_name: str, version: str) -> str:
-    monaihosting_root_path = "https://api.ngc.nvidia.com/v2/models/nvidia/monaihosting"
-    return f"{monaihosting_root_path}/{model_name.lower()}/versions/{version}/files/{model_name}_v{version}.zip"
+    return f"{MONAI_HOSTING_BASE_URL}/{model_name.lower()}/versions/{version}/files/{model_name}_v{version}.zip"
 
 
 def _download_from_github(repo: str, download_path: Path, filename: str, progress: bool = True) -> None:
@@ -267,8 +273,7 @@ def _get_ngc_token(api_key, retry=0):
 
 
 def _get_latest_bundle_version_monaihosting(name):
-    url = "https://api.ngc.nvidia.com/v2/models/nvidia/monaihosting"
-    full_url = f"{url}/{name.lower()}"
+    full_url = f"{MONAI_HOSTING_BASE_URL}/{name.lower()}"
     requests_get, has_requests = optional_import("requests", name="get")
     if has_requests:
         resp = requests_get(full_url)
@@ -279,36 +284,114 @@ def _get_latest_bundle_version_monaihosting(name):
     return model_info["model"]["latestVersionIdStr"]
 
 
-def _get_latest_bundle_version_private_registry(name, repo, headers=None):
-    url = f"https://api.ngc.nvidia.com/v2/{repo}/models"
-    full_url = f"{url}/{name.lower()}"
-    requests_get, has_requests = optional_import("requests", name="get")
-    if has_requests:
-        headers = {} if headers is None else headers
-        resp = requests_get(full_url, headers=headers)
-        resp.raise_for_status()
-    else:
-        raise ValueError("NGC API requires requests package. Please install it.")
+def _examine_monai_version(monai_version: str) -> tuple[bool, str]:
+    """Examine if the package version is compatible with the MONAI version in the metadata."""
+    version_dict = get_versions()
+    package_version = version_dict.get("version", "0+unknown")
+    if package_version == "0+unknown":
+        return False, "Package version is not available. Skipping version check."
+    if monai_version == "0+unknown":
+        return False, "MONAI version is not specified in the bundle. Skipping version check."
+    # treat rc versions as the same as the release version
+    package_version = re.sub(r"rc\d.*", "", package_version)
+    monai_version = re.sub(r"rc\d.*", "", monai_version)
+    if package_version < monai_version:
+        return (
+            False,
+            f"Your MONAI version is {package_version}, but the bundle is built on MONAI version {monai_version}.",
+        )
+    return True, ""
+
+
+def _check_monai_version(bundle_dir: PathLike, name: str) -> None:
+    """Get the `monai_version` from the metadata.json and compare if it is smaller than the installed `monai` package version"""
+    metadata_file = Path(bundle_dir) / name / "configs" / "metadata.json"
+    if not metadata_file.exists():
+        logger.warning(f"metadata file not found in {metadata_file}.")
+        return
+    with open(metadata_file) as f:
+        metadata = json.load(f)
+    is_compatible, msg = _examine_monai_version(metadata.get("monai_version", "0+unknown"))
+    if not is_compatible:
+        logger.warning(msg)
+
+
+def _list_latest_versions(data: dict, max_versions: int = 3) -> list[str]:
+    """
+    Extract the latest versions from the data dictionary.
+
+    Args:
+        data: the data dictionary.
+        max_versions: the maximum number of versions to return.
+
+    Returns:
+        versions of the latest models in the reverse order of creation date, e.g. ['1.0.0', '0.9.0', '0.8.0'].
+    """
+    # Check if the data is a dictionary and it has the key 'modelVersions'
+    if not isinstance(data, dict) or "modelVersions" not in data:
+        raise ValueError("The data is not a dictionary or it does not have the key 'modelVersions'.")
+
+    # Extract the list of model versions
+    model_versions = data["modelVersions"]
+
+    if (
+        not isinstance(model_versions, list)
+        or len(model_versions) == 0
+        or "createdDate" not in model_versions[0]
+        or "versionId" not in model_versions[0]
+    ):
+        raise ValueError(
+            "The model versions are not a list or it is empty or it does not have the keys 'createdDate' and 'versionId'."
+        )
+
+    # Sort the versions by the 'createdDate' in descending order
+    sorted_versions = sorted(model_versions, key=lambda x: x["createdDate"], reverse=True)
+    return [v["versionId"] for v in sorted_versions[:max_versions]]
+
+
+def _get_latest_bundle_version_ngc(name: str, repo: str | None = None, headers: dict | None = None) -> str:
+    base_url = _get_ngc_private_base_url(repo) if repo else NGC_BASE_URL
+    version_endpoint = base_url + f"/{name.lower()}/versions/"
+
+    if not has_requests:
+        raise ValueError("requests package is required, please install it.")
+
+    version_header = {"Accept-Encoding": "gzip, deflate"}  # Excluding 'zstd' to fit NGC requirements
+    if headers:
+        version_header.update(headers)
+    resp = requests_get(version_endpoint, headers=version_header)
+    resp.raise_for_status()
     model_info = json.loads(resp.text)
-    return model_info["model"]["latestVersionIdStr"]
+    latest_versions = _list_latest_versions(model_info)
+
+    for version in latest_versions:
+        file_endpoint = base_url + f"/{name.lower()}/versions/{version}/files/configs/metadata.json"
+        resp = requests_get(file_endpoint, headers=headers)
+        metadata = json.loads(resp.text)
+        resp.raise_for_status()
+        # if the package version is not available or the model is compatible with the package version
+        is_compatible, _ = _examine_monai_version(metadata["monai_version"])
+        if is_compatible:
+            if version != latest_versions[0]:
+                logger.info(f"Latest version is {latest_versions[0]}, but the compatible version is {version}.")
+            return version
+
+    # if no compatible version is found, return the latest version
+    return latest_versions[0]
 
 
 def _get_latest_bundle_version(
     source: str, name: str, repo: str, **kwargs: Any
 ) -> dict[str, list[str] | str] | Any | None:
     if source == "ngc":
         name = _add_ngc_prefix(name)
-        model_dict = _get_all_ngc_models(name)
-        for v in model_dict.values():
-            if v["name"] == name:
-                return v["latest"]
-        return None
+        return _get_latest_bundle_version_ngc(name)
     elif source == "monaihosting":
         return _get_latest_bundle_version_monaihosting(name)
     elif source == "ngc_private":
         headers = kwargs.pop("headers", {})
         name = _add_ngc_prefix(name)
-        return _get_latest_bundle_version_private_registry(name, repo, headers)
+        return _get_latest_bundle_version_ngc(name, repo=repo, headers=headers)
     elif source == "github":
         repo_owner, repo_name, tag_name = repo.split("/")
         return get_bundle_versions(name, repo=f"{repo_owner}/{repo_name}", tag=tag_name)["latest_version"]
@@ -470,9 +553,8 @@ def download(
         if version_ is None:
             version_ = _get_latest_bundle_version(source=source_, name=name_, repo=repo_, headers=headers)
         if source_ == "github":
-            if version_ is not None:
-                name_ = "_v".join([name_, version_])
-            _download_from_github(repo=repo_, download_path=bundle_dir_, filename=name_, progress=progress_)
+            name_ver = "_v".join([name_, version_]) if version_ is not None else name_
+            _download_from_github(repo=repo_, download_path=bundle_dir_, filename=name_ver, progress=progress_)
         elif source_ == "monaihosting":
             _download_from_monaihosting(download_path=bundle_dir_, filename=name_, version=version_, progress=progress_)
         elif source_ == "ngc":
@@ -501,6 +583,8 @@ def download(
                 f"got source: {source_}."
             )
 
+    _check_monai_version(bundle_dir_, name_)
+
 
 @deprecated_arg("net_name", since="1.2", removed="1.5", msg_suffix="please use ``model`` instead.")
 @deprecated_arg("net_kwargs", since="1.2", removed="1.5", msg_suffix="please use ``model`` instead.")

monai/data/test_time_augmentation.py

@@ -106,6 +106,8 @@ class TestTimeAugmentation:
             mode, mean, std, vvc = tt_aug(test_data)
     """
 
+    __test__ = False  # indicate to pytest that this class is not intended for collection
+
     def __init__(
         self,
         transform: InvertibleTransform,

monai/metrics/cumulative_average.py

@@ -65,6 +65,7 @@ def get_current(self, to_numpy: bool = True) -> NdarrayOrTensor:
         if self.val is None:
             return 0
 
+        val: NdarrayOrTensor
         val = self.val.clone()
         val[~torch.isfinite(val)] = 0
 
@@ -96,6 +97,7 @@ def aggregate(self, to_numpy: bool = True) -> NdarrayOrTensor:
             dist.all_reduce(sum)
             dist.all_reduce(count)
 
+        val: NdarrayOrTensor
         val = torch.where(count > 0, sum / count, sum)
 
         if to_numpy:

monai/metrics/panoptic_quality.py

@@ -274,7 +274,7 @@ def _get_paired_iou(
 
         return paired_iou, paired_true, paired_pred
 
-    pairwise_iou = pairwise_iou.cpu().numpy()
+    pairwise_iou = pairwise_iou.cpu().numpy()  # type: ignore[assignment]
     paired_true, paired_pred = linear_sum_assignment(-pairwise_iou)
     paired_iou = pairwise_iou[paired_true, paired_pred]
     paired_true = torch.as_tensor(list(paired_true[paired_iou > match_iou_threshold] + 1), device=device)

monai/metrics/rocauc.py

@@ -88,8 +88,8 @@ def _calculate(y_pred: torch.Tensor, y: torch.Tensor) -> float:
 
     n = len(y)
     indices = y_pred.argsort()
-    y = y[indices].cpu().numpy()
-    y_pred = y_pred[indices].cpu().numpy()
+    y = y[indices].cpu().numpy()  # type: ignore[assignment]
+    y_pred = y_pred[indices].cpu().numpy()  # type: ignore[assignment]
     nneg = auc = tmp_pos = tmp_neg = 0.0
 
     for i in range(n):

monai/transforms/croppad/functional.py

@@ -48,7 +48,7 @@ def _np_pad(img: NdarrayTensor, pad_width: list[tuple[int, int]], mode: str, **k
             warnings.warn(f"Padding: moving img {img.shape} from cuda to cpu for dtype={img.dtype} mode={mode}.")
         img_np = img.detach().cpu().numpy()
     else:
-        img_np = img
+        img_np = np.asarray(img)
     mode = convert_pad_mode(dst=img_np, mode=mode).value
     if mode == "constant" and "value" in kwargs:
         kwargs["constant_values"] = kwargs.pop("value")

monai/visualize/img2tensorboard.py

@@ -176,7 +176,9 @@ def plot_2d_or_3d_image(
     # as the `d` data has no batch dim, reduce the spatial dim index if positive
     frame_dim = frame_dim - 1 if frame_dim > 0 else frame_dim
 
-    d: np.ndarray = data_index.detach().cpu().numpy() if isinstance(data_index, torch.Tensor) else data_index
+    d: np.ndarray = (
+        data_index.detach().cpu().numpy() if isinstance(data_index, torch.Tensor) else np.asarray(data_index)
+    )
 
     if d.ndim == 2:
         d = rescale_array(d, 0, 1)  # type: ignore

tests/test_arraydataset.py

@@ -40,6 +40,7 @@
 
 
 class TestCompose(Compose):
+    __test__ = False  # indicate to pytest that this class is not intended for collection
 
     def __call__(self, input_, lazy=False):
         img = self.transforms[0](input_)

tests/test_auto3dseg.py

@@ -123,6 +123,8 @@ class TestOperations(Operations):
     Test example for user operation
     """
 
+    __test__ = False  # indicate to pytest that this class is not intended for collection
+
     def __init__(self) -> None:
         self.data = {"max": np.max, "mean": np.mean, "min": np.min}
 
@@ -132,6 +134,8 @@ class TestAnalyzer(Analyzer):
     Test example for a simple Analyzer
     """
 
+    __test__ = False  # indicate to pytest that this class is not intended for collection
+
     def __init__(self, key, report_format, stats_name="test"):
         self.key = key
         super().__init__(stats_name, report_format)
@@ -149,6 +153,8 @@ class TestImageAnalyzer(Analyzer):
     Test example for a simple Analyzer
     """
 
+    __test__ = False  # indicate to pytest that this class is not intended for collection
+
     def __init__(self, image_key="image", stats_name="test_image"):
         self.image_key = image_key
         report_format = {"test_stats": None}