Merge branch 'dev' into dev

.github/workflows/cron.yml

@@ -13,24 +13,24 @@ jobs:
     strategy:
       matrix:
         environment:
-          - "PT191+CUDA113"
           - "PT110+CUDA113"
-          - "PT113+CUDA113"
-          - "PTLATEST+CUDA121"
+          - "PT113+CUDA118"
+          - "PT210+CUDA121"
+          - "PTLATEST+CUDA124"
         include:
           # https://docs.nvidia.com/deeplearning/frameworks/pytorch-release-notes
           - environment: PT110+CUDA113
             pytorch: "torch==1.10.2 torchvision==0.11.3 --extra-index-url https://download.pytorch.org/whl/cu113"
             base: "nvcr.io/nvidia/pytorch:21.06-py3"  # CUDA 11.3
-          - environment: PT113+CUDA113
-            pytorch: "torch==1.13.1 torchvision==0.14.1 --extra-index-url https://download.pytorch.org/whl/cu113"
-            base: "nvcr.io/nvidia/pytorch:21.06-py3"  # CUDA 11.3
-          - environment: PT113+CUDA122
+          - environment: PT113+CUDA118
             pytorch: "torch==1.13.1 torchvision==0.14.1 --extra-index-url https://download.pytorch.org/whl/cu121"
-            base: "nvcr.io/nvidia/pytorch:23.08-py3"  # CUDA 12.2
+            base: "nvcr.io/nvidia/pytorch:22.10-py3"  # CUDA 11.8
+          - environment: PT210+CUDA121
+            pytorch: "pytorch==2.1.0 torchvision==0.16.0 --extra-index-url https://download.pytorch.org/whl/cu121"
+            base: "nvcr.io/nvidia/pytorch:23.08-py3"  # CUDA 12.1
           - environment: PTLATEST+CUDA124
             pytorch: "-U torch torchvision --extra-index-url https://download.pytorch.org/whl/cu121"
-            base: "nvcr.io/nvidia/pytorch:24.03-py3"  # CUDA 12.4
+            base: "nvcr.io/nvidia/pytorch:24.08-py3"  # CUDA 12.4
     container:
       image: ${{ matrix.base }}
       options: "--gpus all"
@@ -80,7 +80,7 @@ jobs:
     if: github.repository == 'Project-MONAI/MONAI'
     strategy:
       matrix:
-        container: ["pytorch:23.08", "pytorch:24.03"]
+        container: ["pytorch:23.08", "pytorch:24.08"]
     container:
       image: nvcr.io/nvidia/${{ matrix.container }}-py3  # testing with the latest pytorch base image
       options: "--gpus all"
@@ -129,7 +129,7 @@ jobs:
     if: github.repository == 'Project-MONAI/MONAI'
     strategy:
       matrix:
-        container: ["pytorch:24.03"]
+        container: ["pytorch:24.08"]
     container:
       image: nvcr.io/nvidia/${{ matrix.container }}-py3  # testing with the latest pytorch base image
       options: "--gpus all"
@@ -233,7 +233,7 @@ jobs:
     if: github.repository == 'Project-MONAI/MONAI'
     needs: cron-gpu  # so that monai itself is verified first
     container:
-      image: nvcr.io/nvidia/pytorch:24.03-py3  # testing with the latest pytorch base image
+      image: nvcr.io/nvidia/pytorch:24.08-py3  # testing with the latest pytorch base image
       options: "--gpus all --ipc=host"
     runs-on: [self-hosted, linux, x64, integration]
     steps:

.github/workflows/pythonapp-gpu.yml

@@ -44,9 +44,9 @@ jobs:
             pytorch: "-h"  # we explicitly set pytorch to -h to avoid pip install error
             base: "nvcr.io/nvidia/pytorch:23.08-py3"
           - environment: PT210+CUDA121DOCKER
-            # 24.03: 2.3.0a0+40ec155e58.nv24.3
+            # 24.08: 2.3.0a0+40ec155e58.nv24.3
             pytorch: "-h"  # we explicitly set pytorch to -h to avoid pip install error
-            base: "nvcr.io/nvidia/pytorch:24.03-py3"
+            base: "nvcr.io/nvidia/pytorch:24.08-py3"
     container:
       image: ${{ matrix.base }}
       options: --gpus all --env NVIDIA_DISABLE_REQUIRE=true  # workaround for unsatisfied condition: cuda>=11.6

Dockerfile

@@ -11,7 +11,7 @@
 
 # To build with a different base image
 # please run `docker build` using the `--build-arg PYTORCH_IMAGE=...` flag.
-ARG PYTORCH_IMAGE=nvcr.io/nvidia/pytorch:24.03-py3
+ARG PYTORCH_IMAGE=nvcr.io/nvidia/pytorch:24.08-py3
 FROM ${PYTORCH_IMAGE}
 
 LABEL maintainer="[email protected]"
@@ -56,4 +56,5 @@ RUN apt-get update \
   && rm -rf /var/lib/apt/lists/*
 # append /opt/tools to runtime path for NGC CLI to be accessible from all file system locations
 ENV PATH=${PATH}:/opt/tools
+ENV POLYGRAPHY_AUTOINSTALL_DEPS=1
 WORKDIR /opt/monai

docs/requirements.txt

@@ -42,3 +42,4 @@ zarr
 huggingface_hub
 pyamg>=5.0.0
 packaging
+polygraphy

docs/source/config_syntax.md

@@ -16,6 +16,7 @@ Content:
   - [`$` to evaluate as Python expressions](#to-evaluate-as-python-expressions)
   - [`%` to textually replace configuration elements](#to-textually-replace-configuration-elements)
   - [`_target_` (`_disabled_`, `_desc_`, `_requires_`, `_mode_`) to instantiate a Python object](#instantiate-a-python-object)
+  - [`+` to alter semantics of merging config keys from multiple configuration files](#multiple-config-files)
 - [The command line interface](#the-command-line-interface)
 - [Recommendations](#recommendations)
 
@@ -175,6 +176,47 @@ _Description:_ `_requires_`, `_disabled_`, `_desc_`, and `_mode_` are optional k
   - `"debug"` -- execute with debug prompt and return the return value of ``pdb.runcall(_target_, **kwargs)``,
     see also [`pdb.runcall`](https://docs.python.org/3/library/pdb.html#pdb.runcall).
 
+## Multiple config files
+
+_Description:_ Multiple config files may be specified on the command line.
+The content of those config files is being merged. When same keys are specifiled in more than one config file,
+the value associated with the key is being overridden, in the order config files are specified.
+If the desired behaviour is to merge values from both files, the key in second config file should be prefixed with `+`.
+The value types for the merged contents must match and be both of `dict` or both of `list` type.
+`dict` values will be merged via update(), `list` values - concatenated via extend().
+Here's an example. In this case, "amp" value will be overridden by extra_config.json.
+`imports` and `preprocessing#transforms` lists will be merged. An error would be thrown if the value type in `"+imports"` is not `list`:
+
+config.json:
+```json
+{
+    "amp": "$True"
+    "imports": [
+	"$import torch"
+    ],
+    "preprocessing": {
+        "_target_": "Compose",
+        "transforms": [
+	  "$@t1",
+	  "$@t2"
+        ]
+    },
+}
+```
+
+extra_config.json:
+```json
+{
+    "amp": "$False"
+    "+imports": [
+	"$from monai.networks import trt_compile"
+    ],
+    "+preprocessing#transforms": [
+        "$@t3"
+    ]
+}
+```
+
 ## The command line interface
 
 In addition to the Pythonic APIs, a few command line interfaces (CLI) are provided to interact with the bundle.

docs/source/transforms.rst

@@ -554,6 +554,12 @@ IO
     :members:
     :special-members: __call__
 
+`WriteFileMapping`
+""""""""""""""""""
+.. autoclass:: WriteFileMapping
+    :members:
+    :special-members: __call__
+
 
 NVIDIA Tool Extension (NVTX)
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
@@ -970,6 +976,18 @@ Spatial
     :members:
     :special-members: __call__
 
+`ConvertBoxToPoints`
+""""""""""""""""""""
+.. autoclass:: ConvertBoxToPoints
+    :members:
+    :special-members: __call__
+
+`ConvertPointsToBoxes`
+""""""""""""""""""""""
+.. autoclass:: ConvertPointsToBoxes
+    :members:
+    :special-members: __call__
+
 
 Smooth Field
 ^^^^^^^^^^^^
@@ -1216,6 +1234,12 @@ Utility
     :members:
     :special-members: __call__
 
+`ApplyTransformToPoints`
+""""""""""""""""""""""""
+.. autoclass:: ApplyTransformToPoints
+    :members:
+    :special-members: __call__
+
 Dictionary Transforms
 ---------------------
 
@@ -1642,6 +1666,12 @@ IO (Dict)
     :members:
     :special-members: __call__
 
+`WriteFileMappingd`
+"""""""""""""""""""
+.. autoclass:: WriteFileMappingd
+    :members:
+    :special-members: __call__
+
 Post-processing (Dict)
 ^^^^^^^^^^^^^^^^^^^^^^
 
@@ -1961,6 +1991,18 @@ Spatial (Dict)
     :members:
     :special-members: __call__
 
+`ConvertBoxToPointsd`
+"""""""""""""""""""""
+.. autoclass:: ConvertBoxToPointsd
+    :members:
+    :special-members: __call__
+
+`ConvertPointsToBoxesd`
+"""""""""""""""""""""""
+.. autoclass:: ConvertPointsToBoxesd
+    :members:
+    :special-members: __call__
+
 
 Smooth Field (Dict)
 ^^^^^^^^^^^^^^^^^^^
@@ -2265,6 +2307,12 @@ Utility (Dict)
     :members:
     :special-members: __call__
 
+`ApplyTransformToPointsd`
+"""""""""""""""""""""""""
+.. autoclass:: ApplyTransformToPointsd
+    :members:
+    :special-members: __call__
+
 
 MetaTensor
 ^^^^^^^^^^

monai/apps/vista3d/inferer.py

@@ -100,7 +100,7 @@ def point_based_window_inferer(
                         point_labels=point_labels,
                         class_vector=class_vector,
                         prompt_class=prompt_class,
-                        patch_coords=unravel_slice,
+                        patch_coords=[unravel_slice],
                         prev_mask=prev_mask,
                         **kwargs,
                     )

monai/apps/vista3d/sampler.py

@@ -20,8 +20,6 @@
 import torch
 from torch import Tensor
 
-__all__ = ["sample_prompt_pairs"]
-
 ENABLE_SPECIAL = True
 SPECIAL_INDEX = (23, 24, 25, 26, 27, 57, 128)
 MERGE_LIST = {
@@ -30,6 +28,8 @@
     132: [57],  # overlap with trachea merge into airway
 }
 
+__all__ = ["sample_prompt_pairs"]
+
 
 def _get_point_label(id: int) -> tuple[int, int]:
     if id in SPECIAL_INDEX and ENABLE_SPECIAL:
@@ -66,22 +66,29 @@ def sample_prompt_pairs(
         max_backprompt: int, max number of prompt from background.
         max_point: maximum number of points for each object.
         include_background: if include 0 into training prompt. If included, background 0 is treated
-            the same as foreground. Always be False for multi-partial-dataset training. If needed,
-            can be true for finetuning specific dataset, .
+            the same as foreground and points will be sampled. Can be true only if user want to segment
+            background 0 with point clicks, otherwise always be false.
         drop_label_prob: probability to drop label prompt.
         drop_point_prob: probability to drop point prompt.
         point_sampler: sampler to augment masks with supervoxel.
         point_sampler_kwargs: arguments for point_sampler.
 
     Returns:
-        label_prompt: [B, 1]. The classes used for training automatic segmentation.
-        point: [B, N, 3]. The corresponding points for each class.
-        Note that background label prompt requires matching point as well ([0,0,0] is used).
-        point_label: [B, N]. The corresponding point labels for each point (negative or positive).
-        -1 is used for padding the background label prompt and will be ignored.
-        prompt_class: [B, 1], exactly the same with label_prompt for label indexing for training loss.
-        label_prompt can be None, and prompt_class is used to identify point classes.
+        tuple:
+            - label_prompt (Tensor | None): Tensor of shape [B, 1] containing the classes used for
+              training automatic segmentation.
+            - point (Tensor | None): Tensor of shape [B, N, 3] representing the corresponding points
+              for each class. Note that background label prompts require matching points as well
+              (e.g., [0, 0, 0] is used).
+            - point_label (Tensor | None): Tensor of shape [B, N] representing the corresponding point
+              labels for each point (negative or positive). -1 is used for padding the background
+              label prompt and will be ignored.
+            - prompt_class (Tensor | None): Tensor of shape [B, 1], exactly the same as label_prompt
+              for label indexing during training. If label_prompt is None, prompt_class is used to
+              identify point classes.
+
     """
+
     # class label number
     if not labels.shape[0] == 1:
         raise ValueError("only support batch size 1")

monai/bundle/config_parser.py

@@ -20,7 +20,7 @@
 
 from monai.bundle.config_item import ComponentLocator, ConfigComponent, ConfigExpression, ConfigItem
 from monai.bundle.reference_resolver import ReferenceResolver
-from monai.bundle.utils import ID_REF_KEY, ID_SEP_KEY, MACRO_KEY
+from monai.bundle.utils import ID_REF_KEY, ID_SEP_KEY, MACRO_KEY, merge_kv
 from monai.config import PathLike
 from monai.utils import ensure_tuple, look_up_option, optional_import
 from monai.utils.misc import CheckKeyDuplicatesYamlLoader, check_key_duplicates
@@ -423,8 +423,10 @@ def load_config_files(cls, files: PathLike | Sequence[PathLike] | dict, **kwargs
         if isinstance(files, str) and not Path(files).is_file() and "," in files:
             files = files.split(",")
         for i in ensure_tuple(files):
-            for k, v in (cls.load_config_file(i, **kwargs)).items():
-                parser[k] = v
+            config_dict = cls.load_config_file(i, **kwargs)
+            for k, v in config_dict.items():
+                merge_kv(parser, k, v)
+
         return parser.get()  # type: ignore
 
     @classmethod

monai/bundle/scripts.py

@@ -32,7 +32,7 @@
 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
-from monai.bundle.utils import DEFAULT_INFERENCE, DEFAULT_METADATA
+from monai.bundle.utils import DEFAULT_INFERENCE, DEFAULT_METADATA, merge_kv
 from monai.bundle.workflows import BundleWorkflow, ConfigWorkflow
 from monai.config import IgniteInfo, PathLike
 from monai.data import load_net_with_metadata, save_net_with_metadata
@@ -105,7 +105,7 @@ def update_kwargs(args: str | dict | None = None, ignore_none: bool = True, **kw
         if isinstance(v, dict) and isinstance(args_.get(k), dict):
             args_[k] = update_kwargs(args_[k], ignore_none, **v)
         else:
-            args_[k] = v
+            merge_kv(args_, k, v)
     return args_