merge master

monai/config/__init__.py

@@ -18,4 +18,4 @@
     print_gpu_info,
     print_system_info,
 )
-from .type_definitions import IndexSelection, KeysCollection
+from .type_definitions import DtypeLike, IndexSelection, KeysCollection, NdarrayTensor

monai/config/deviceconfig.py

@@ -162,7 +162,7 @@ def get_system_info() -> OrderedDict:
             _dict_append(
                 output,
                 "Avg. sensor temp. (Celsius)",
-                lambda: round(
+                lambda: np.round(
                     np.mean([item.current for sublist in psutil.sensors_temperatures().values() for item in sublist], 1)
                 ),
             )

monai/config/type_definitions.py

@@ -9,9 +9,12 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from typing import Collection, Hashable, Iterable, Union
+from typing import Collection, Hashable, Iterable, TypeVar, Union
 
-__all__ = ["KeysCollection", "IndexSelection"]
+import numpy as np
+import torch
+
+__all__ = ["KeysCollection", "IndexSelection", "DtypeLike", "NdarrayTensor"]
 
 """Commonly used concepts
 This module provides naming and type specifications for commonly used concepts
@@ -51,3 +54,16 @@
 The indices must be integers, and if a container of indices is specified, the
 container must be iterable.
 """
+
+DtypeLike = Union[
+    np.dtype,
+    type,
+    None,
+]
+"""Type of datatypes
+adapted from https://github.com/numpy/numpy/blob/master/numpy/typing/_dtype_like.py
+"""
+
+# Generic type which can represent either a numpy.ndarray or a torch.Tensor
+# Unlike Union can create a dependence between parameter(s) / return(s)
+NdarrayTensor = TypeVar("NdarrayTensor", np.ndarray, torch.Tensor)

monai/data/csv_saver.py

@@ -17,6 +17,8 @@
 import numpy as np
 import torch
 
+from monai.utils import ImageMetaKey as Key
+
 
 class CSVSaver:
     """
@@ -73,9 +75,9 @@ def save(self, data: Union[torch.Tensor, np.ndarray], meta_data: Optional[Dict]
             meta_data: the meta data information corresponding to the data.
 
         """
-        save_key = meta_data["filename_or_obj"] if meta_data else str(self._data_index)
+        save_key = meta_data[Key.FILENAME_OR_OBJ] if meta_data else str(self._data_index)
         self._data_index += 1
-        if torch.is_tensor(data):
+        if isinstance(data, torch.Tensor):
             data = data.detach().cpu().numpy()
         if not isinstance(data, np.ndarray):
             raise AssertionError

monai/data/dataset.py

@@ -571,6 +571,10 @@ class SmartCacheDataset(CacheDataset):
         3. Call `update_cache()` before every epoch to replace training items.
         4. Call `shutdown()` when training ends.
 
+    Note:
+        This replacement will not work if setting the `multiprocessing_context` of DataLoader to `spawn`
+        or on windows(the default multiprocessing method is `spawn`) and setting `num_workers` greater than 0.
+
     """
 
     def __init__(
@@ -601,7 +605,7 @@ def __init__(
         if self._cache is None:
             self._cache = self._fill_cache()
         if self.cache_num >= len(data):
-            raise ValueError("cache_num must be smaller than dataset length to support replacement.")
+            warnings.warn("cache_num is greater or equal than dataset length, fall back to regular CacheDataset.")
         if replace_rate <= 0:
             raise ValueError("replace_rate must be greater than 0, otherwise, please use CacheDataset.")
         self.num_replace_workers: int = num_replace_workers
@@ -637,7 +641,7 @@ def is_started(self):
         """
         if self._replace_mgr is None:
             return False
-        return self._replace_mgr.isAlive()
+        return self._replace_mgr.is_alive()
 
     def start(self):
         """
@@ -688,7 +692,7 @@ def update_cache(self):
         If the cache has been shutdown before, need to restart the `_replace_mgr` thread.
 
         """
-        if not self._replace_mgr.isAlive():
+        if not self._replace_mgr.is_alive():
             self._restart()
 
         # make sure update is done

monai/data/image_dataset.py

@@ -14,6 +14,7 @@
 import numpy as np
 from torch.utils.data import Dataset
 
+from monai.config import DtypeLike
 from monai.data.image_reader import ImageReader
 from monai.transforms import LoadImage, Randomizable, apply_transform
 from monai.utils import MAX_SEED, get_seed
@@ -36,7 +37,7 @@ def __init__(
         transform: Optional[Callable] = None,
         seg_transform: Optional[Callable] = None,
         image_only: bool = True,
-        dtype: Optional[np.dtype] = np.float32,
+        dtype: DtypeLike = np.float32,
         reader: Optional[Union[ImageReader, str]] = None,
         *args,
         **kwargs,

monai/data/image_reader.py

@@ -16,7 +16,7 @@
 import numpy as np
 from torch.utils.data._utils.collate import np_str_obj_array_pattern
 
-from monai.config import KeysCollection
+from monai.config import DtypeLike, KeysCollection
 from monai.data.utils import correct_nifti_header_if_necessary
 from monai.utils import ensure_tuple, optional_import
 
@@ -244,7 +244,7 @@ def _get_affine(self, img) -> np.ndarray:
         affine = np.eye(direction.shape[0] + 1)
         affine[(slice(-1), slice(-1))] = direction @ np.diag(spacing)
         affine[(slice(-1), -1)] = origin
-        return affine
+        return np.asarray(affine)
 
     def _get_spatial_shape(self, img) -> np.ndarray:
         """
@@ -258,7 +258,7 @@ def _get_spatial_shape(self, img) -> np.ndarray:
         shape.reverse()
         return np.asarray(shape)
 
-    def _get_array_data(self, img) -> np.ndarray:
+    def _get_array_data(self, img):
         """
         Get the raw array data of the image, converted to Numpy array.
 
@@ -295,7 +295,7 @@ class NibabelReader(ImageReader):
 
     """
 
-    def __init__(self, as_closest_canonical: bool = False, dtype: Optional[np.dtype] = np.float32, **kwargs):
+    def __init__(self, as_closest_canonical: bool = False, dtype: DtypeLike = np.float32, **kwargs):
         super().__init__()
         self.as_closest_canonical = as_closest_canonical
         self.dtype = dtype
@@ -385,7 +385,7 @@ def _get_affine(self, img) -> np.ndarray:
             img: a Nibabel image object loaded from a image file.
 
         """
-        return img.affine.copy()
+        return np.array(img.affine, copy=True)
 
     def _get_spatial_shape(self, img) -> np.ndarray:
         """

monai/data/nifti_saver.py

@@ -14,9 +14,11 @@
 import numpy as np
 import torch
 
+from monai.config import DtypeLike
 from monai.data.nifti_writer import write_nifti
 from monai.data.utils import create_file_basename
 from monai.utils import GridSampleMode, GridSamplePadMode
+from monai.utils import ImageMetaKey as Key
 
 
 class NiftiSaver:
@@ -36,8 +38,8 @@ def __init__(
         mode: Union[GridSampleMode, str] = GridSampleMode.BILINEAR,
         padding_mode: Union[GridSamplePadMode, str] = GridSamplePadMode.BORDER,
         align_corners: bool = False,
-        dtype: Optional[np.dtype] = np.float64,
-        output_dtype: Optional[np.dtype] = np.float32,
+        dtype: DtypeLike = np.float64,
+        output_dtype: DtypeLike = np.float32,
     ) -> None:
         """
         Args:
@@ -94,13 +96,13 @@ def save(self, data: Union[torch.Tensor, np.ndarray], meta_data: Optional[Dict]
         See Also
             :py:meth:`monai.data.nifti_writer.write_nifti`
         """
-        filename = meta_data["filename_or_obj"] if meta_data else str(self._data_index)
+        filename = meta_data[Key.FILENAME_OR_OBJ] if meta_data else str(self._data_index)
         self._data_index += 1
         original_affine = meta_data.get("original_affine", None) if meta_data else None
         affine = meta_data.get("affine", None) if meta_data else None
         spatial_shape = meta_data.get("spatial_shape", None) if meta_data else None
 
-        if torch.is_tensor(data):
+        if isinstance(data, torch.Tensor):
             data = data.detach().cpu().numpy()
 
         filename = create_file_basename(self.output_postfix, filename, self.output_dir)
@@ -109,7 +111,7 @@ def save(self, data: Union[torch.Tensor, np.ndarray], meta_data: Optional[Dict]
         while len(data.shape) < 4:
             data = np.expand_dims(data, -1)
         # change data to "channel last" format and write to nifti format file
-        data = np.moveaxis(data, 0, -1)
+        data = np.moveaxis(np.asarray(data), 0, -1)
         write_nifti(
             data,
             file_name=filename,

monai/data/nifti_writer.py

@@ -14,6 +14,7 @@
 import numpy as np
 import torch
 
+from monai.config import DtypeLike
 from monai.data.utils import compute_shape_offset, to_affine_nd
 from monai.networks.layers import AffineTransform
 from monai.utils import GridSampleMode, GridSamplePadMode, optional_import
@@ -27,12 +28,12 @@ def write_nifti(
     affine: Optional[np.ndarray] = None,
     target_affine: Optional[np.ndarray] = None,
     resample: bool = True,
-    output_spatial_shape: Optional[Sequence[int]] = None,
+    output_spatial_shape: Union[Sequence[int], np.ndarray, None] = None,
     mode: Union[GridSampleMode, str] = GridSampleMode.BILINEAR,
     padding_mode: Union[GridSamplePadMode, str] = GridSamplePadMode.BORDER,
     align_corners: bool = False,
-    dtype: Optional[np.dtype] = np.float64,
-    output_dtype: Optional[np.dtype] = np.float32,
+    dtype: DtypeLike = np.float64,
+    output_dtype: DtypeLike = np.float32,
 ) -> None:
     """
     Write numpy data into NIfTI files to disk.  This function converts data
@@ -126,7 +127,7 @@ def write_nifti(
     transform = np.linalg.inv(_affine) @ target_affine
     if output_spatial_shape is None:
         output_spatial_shape, _ = compute_shape_offset(data.shape, _affine, target_affine)
-    output_spatial_shape_ = list(output_spatial_shape)
+    output_spatial_shape_ = list(output_spatial_shape) if output_spatial_shape is not None else []
     if data.ndim > 3:  # multi channel, resampling each channel
         while len(output_spatial_shape_) < 3:
             output_spatial_shape_ = output_spatial_shape_ + [1]

monai/data/png_saver.py

@@ -16,6 +16,7 @@
 
 from monai.data.png_writer import write_png
 from monai.data.utils import create_file_basename
+from monai.utils import ImageMetaKey as Key
 from monai.utils import InterpolateMode
 
 
@@ -82,11 +83,11 @@ def save(self, data: Union[torch.Tensor, np.ndarray], meta_data: Optional[Dict]
             :py:meth:`monai.data.png_writer.write_png`
 
         """
-        filename = meta_data["filename_or_obj"] if meta_data else str(self._data_index)
+        filename = meta_data[Key.FILENAME_OR_OBJ] if meta_data else str(self._data_index)
         self._data_index += 1
         spatial_shape = meta_data.get("spatial_shape", None) if meta_data and self.resample else None
 
-        if torch.is_tensor(data):
+        if isinstance(data, torch.Tensor):
             data = data.detach().cpu().numpy()
 
         filename = create_file_basename(self.output_postfix, filename, self.output_dir)
@@ -95,12 +96,12 @@ def save(self, data: Union[torch.Tensor, np.ndarray], meta_data: Optional[Dict]
         if data.shape[0] == 1:
             data = data.squeeze(0)
         elif 2 < data.shape[0] < 5:
-            data = np.moveaxis(data, 0, -1)
+            data = np.moveaxis(np.asarray(data), 0, -1)
         else:
             raise ValueError(f"Unsupported number of channels: {data.shape[0]}, available options are [1, 3, 4]")
 
         write_png(
-            data,
+            np.asarray(data),
             file_name=filename,
             output_spatial_shape=spatial_shape,
             mode=self.mode,

monai/data/png_writer.py

@@ -65,10 +65,10 @@ def write_png(
             data = np.expand_dims(data, 0)  # make a channel
             data = xform(data)[0]  # first channel
         if mode != InterpolateMode.NEAREST:
-            data = np.clip(data, _min, _max)
+            data = np.clip(data, _min, _max)  # type: ignore
 
     if scale is not None:
-        data = np.clip(data, 0.0, 1.0)  # png writer only can scale data in range [0, 1]
+        data = np.clip(data, 0.0, 1.0)  # type: ignore # png writer only can scale data in range [0, 1]
         if scale == np.iinfo(np.uint8).max:
             data = (scale * data).astype(np.uint8)
         elif scale == np.iinfo(np.uint16).max:

monai/data/utils.py

@@ -329,7 +329,7 @@ def rectify_header_sform_qform(img_nii):
     return img_nii
 
 
-def zoom_affine(affine: np.ndarray, scale: Sequence[float], diagonal: bool = True) -> np.ndarray:
+def zoom_affine(affine: np.ndarray, scale: Sequence[float], diagonal: bool = True):
     """
     To make column norm of `affine` the same as `scale`.  If diagonal is False,
     returns an affine that combines orthogonal rotation and the new scale.
@@ -379,7 +379,7 @@ def zoom_affine(affine: np.ndarray, scale: Sequence[float], diagonal: bool = Tru
 
 
 def compute_shape_offset(
-    spatial_shape: np.ndarray, in_affine: np.ndarray, out_affine: np.ndarray
+    spatial_shape: Union[np.ndarray, Sequence[int]], in_affine: np.ndarray, out_affine: np.ndarray
 ) -> Tuple[np.ndarray, np.ndarray]:
     """
     Given input and output affine, compute appropriate shapes

monai/engines/utils.py

@@ -32,7 +32,7 @@
 
 class IterationEvents(EventEnum):
     """
-    Addtional Events engine can register and trigger in the iteration process.
+    Additional Events engine can register and trigger in the iteration process.
     Refer to the example in ignite: https://github.com/pytorch/ignite/blob/master/ignite/engine/events.py#L146
     These Events can be triggered during training iteration:
     `FORWARD_COMPLETED` is the Event when `network(image, label)` completed.

monai/handlers/__init__.py

@@ -25,5 +25,11 @@
 from .stats_handler import StatsHandler
 from .surface_distance import SurfaceDistance
 from .tensorboard_handlers import TensorBoardImageHandler, TensorBoardStatsHandler
-from .utils import evenly_divisible_all_gather, stopping_fn_from_loss, stopping_fn_from_metric, write_metrics_reports
+from .utils import (
+    evenly_divisible_all_gather,
+    stopping_fn_from_loss,
+    stopping_fn_from_metric,
+    string_list_all_gather,
+    write_metrics_reports,
+)
 from .validation_handler import ValidationHandler