Merge branch 'dev' into bump-torch-minimum

Signed-off-by: YunLiu <[email protected]>

Author: KumoLiu

.pre-commit-config.yaml

@@ -66,7 +66,7 @@ repos:
           )$
 
   - repo: https://github.com/hadialqattan/pycln
-    rev: v2.4.0
+    rev: v2.5.0
     hooks:
       - id: pycln
         args: [--config=pyproject.toml]

monai/bundle/scripts.py

@@ -174,7 +174,7 @@ 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"{NGC_BASE_URL}/{model_name.lower()}/versions/{version}/zip"
+    return f"{NGC_BASE_URL}/{model_name.lower()}/versions/{version}/files"
 
 
 def _get_ngc_private_base_url(repo: str) -> str:
@@ -218,6 +218,21 @@ def _remove_ngc_prefix(name: str, prefix: str = "monai_") -> str:
     return name
 
 
+def _get_all_download_files(request_url: str, headers: dict | None = None) -> list[str]:
+    if not has_requests:
+        raise ValueError("requests package is required, please install it.")
+    headers = {} if headers is None else headers
+    response = requests_get(request_url, headers=headers)
+    response.raise_for_status()
+    model_info = json.loads(response.text)
+
+    if not isinstance(model_info, dict) or "modelFiles" not in model_info:
+        raise ValueError("The data is not a dictionary or it does not have the key 'modelFiles'.")
+
+    model_files = model_info["modelFiles"]
+    return [f["path"] for f in model_files]
+
+
 def _download_from_ngc(
     download_path: Path,
     filename: str,
@@ -229,12 +244,12 @@ def _download_from_ngc(
     # ensure prefix is contained
     filename = _add_ngc_prefix(filename, prefix=prefix)
     url = _get_ngc_bundle_url(model_name=filename, version=version)
-    filepath = download_path / f"{filename}_v{version}.zip"
     if remove_prefix:
         filename = _remove_ngc_prefix(filename, prefix=remove_prefix)
-    extract_path = download_path / f"{filename}"
-    download_url(url=url, filepath=filepath, hash_val=None, progress=progress)
-    extractall(filepath=filepath, output_dir=extract_path, has_base=True)
+    filepath = download_path / filename
+    filepath.mkdir(parents=True, exist_ok=True)
+    for file in _get_all_download_files(url):
+        download_url(url=f"{url}/{file}", filepath=f"{filepath}/{file}", hash_val=None, progress=progress)
 
 
 def _download_from_ngc_private(

monai/data/image_reader.py

@@ -607,3 +607,8 @@ def print_verbose(self) -> None:
         print(self)
         if self.meta is not None:
             print(self.meta.__repr__())
+
+
+# needed in later versions of Pytorch to indicate the class is safe for serialisation
+if hasattr(torch.serialization, "add_safe_globals"):
+    torch.serialization.add_safe_globals([MetaTensor])

monai/data/meta_tensor.py

@@ -15,12 +15,13 @@
 from abc import ABC, abstractmethod
 from collections.abc import Sequence
 from contextlib import nullcontext
+from tempfile import TemporaryDirectory
 from typing import TYPE_CHECKING, Any
 
 import numpy as np
 import torch
 
-from monai.utils import ensure_tuple_size, optional_import, require_pkg
+from monai.utils import ensure_tuple_size, get_package_version, optional_import, require_pkg, version_geq
 
 if TYPE_CHECKING:
     import zarr
@@ -233,7 +234,7 @@ def __init__(
         store: zarr.storage.Store | str = "merged.zarr",
         value_store: zarr.storage.Store | str | None = None,
         count_store: zarr.storage.Store | str | None = None,
-        compressor: str = "default",
+        compressor: str | None = None,
         value_compressor: str | None = None,
         count_compressor: str | None = None,
         chunks: Sequence[int] | bool = True,
@@ -246,8 +247,22 @@ def __init__(
         self.value_dtype = value_dtype
         self.count_dtype = count_dtype
         self.store = store
-        self.value_store = zarr.storage.TempStore() if value_store is None else value_store
-        self.count_store = zarr.storage.TempStore() if count_store is None else count_store
+        self.tmpdir: TemporaryDirectory | None
+        if version_geq(get_package_version("zarr"), "3.0.0"):
+            if value_store is None:
+                self.tmpdir = TemporaryDirectory()
+                self.value_store = zarr.storage.LocalStore(self.tmpdir.name)
+            else:
+                self.value_store = value_store
+            if count_store is None:
+                self.tmpdir = TemporaryDirectory()
+                self.count_store = zarr.storage.LocalStore(self.tmpdir.name)
+            else:
+                self.count_store = count_store
+        else:
+            self.tmpdir = None
+            self.value_store = zarr.storage.TempStore() if value_store is None else value_store
+            self.count_store = zarr.storage.TempStore() if count_store is None else count_store
         self.chunks = chunks
         self.compressor = compressor
         self.value_compressor = value_compressor

monai/inferers/merger.py

@@ -821,6 +821,7 @@ class NormalizeIntensity(Transform):
     mean and std on each channel separately.
     When `channel_wise` is True, the first dimension of `subtrahend` and `divisor` should
     be the number of image channels if they are not None.
+    If the input is not of floating point type, it will be converted to float32
 
     Args:
         subtrahend: the amount to subtract by (usually the mean).
@@ -907,6 +908,9 @@ def __call__(self, img: NdarrayOrTensor) -> NdarrayOrTensor:
             if self.divisor is not None and len(self.divisor) != len(img):
                 raise ValueError(f"img has {len(img)} channels, but divisor has {len(self.divisor)} components.")
 
+            if not img.dtype.is_floating_point:
+                img, *_ = convert_data_type(img, dtype=torch.float32)
+
             for i, d in enumerate(img):
                 img[i] = self._normalize(  # type: ignore
                     d,

monai/transforms/intensity/array.py

@@ -234,7 +234,7 @@ def plot_engine_status(
 
 
 def _get_loss_from_output(
-    output: list[torch.Tensor | dict[str, torch.Tensor]] | dict[str, torch.Tensor] | torch.Tensor
+    output: list[torch.Tensor | dict[str, torch.Tensor]] | dict[str, torch.Tensor] | torch.Tensor,
 ) -> torch.Tensor:
     """Returns a single value from the network output, which is a dict or tensor."""
 

monai/utils/jupyter_utils.py

@@ -65,11 +65,11 @@ def _image3_animated_gif(
     img_str = b""
     for b_data in PIL.GifImagePlugin.getheader(ims[0])[0]:
         img_str += b_data
-    img_str += b"\x21\xFF\x0B\x4E\x45\x54\x53\x43\x41\x50" b"\x45\x32\x2E\x30\x03\x01\x00\x00\x00"
+    img_str += b"\x21\xff\x0b\x4e\x45\x54\x53\x43\x41\x50" b"\x45\x32\x2e\x30\x03\x01\x00\x00\x00"
     for i in ims:
         for b_data in PIL.GifImagePlugin.getdata(i):
             img_str += b_data
-    img_str += b"\x3B"
+    img_str += b"\x3b"
 
     summary = SummaryX if has_tensorboardx and isinstance(writer, SummaryWriterX) else Summary
     summary_image_str = summary.Image(height=10, width=10, colorspace=1, encoded_image_string=img_str)

monai/visualize/img2tensorboard.py

@@ -18,7 +18,7 @@ pep8-naming
 pycodestyle
 pyflakes
 black>=22.12
-isort>=5.1
+isort>=5.1, <6.0
 ruff
 pytype>=2020.6.1; platform_system != "Windows"
 types-setuptools

requirements-dev.txt

@@ -1,2 +1,2 @@
-torch>=1.13.1
+torch>=1.13.1,<2.6
 numpy>=1.24,<2.0