Merge branch 'code_owners' of github.com:ericspod/MONAI into code_owners

Author: ericspod

.github/CODEOWNERS

@@ -1,4 +1,17 @@
 /monai/ @KumoLiu @ericspod @Nic-Ma
-/docs/ @Project-MONAI/core-reviewers
-/tests/ @Project-MONAI/core-reviewers
+/docs/ @KumoLiu @ericspod @Nic-Ma
+/tests/ @KumoLiu @ericspod @Nic-Ma
 /.github/ @KumoLiu
+/monai/networks/schedulers/ @virginiafdez
+/monai/inferers/inferer.py @virginiafdez
+/monai/losses/adversarial_loss.py @virginiafdez
+/monai/losses/perceptual.py @virginiafdez
+/monai/networks/blocks/spade_norm.py @virginiafdez
+/monai/networks/nets/autoencoderkl.py @virginiafdez
+/monai/networks/nets/controlnet.py @virginiafdez
+/monai/networks/nets/diffusion_model_unet.py @virginiafdez
+/monai/networks/nets/patchgan_discriminator.py @virginiafdez
+/monai/networks/nets/spade_autoencoderkl.py @virginiafdez
+/monai/networks/nets/spade_diffusion_model_unet.py @virginiafdez
+/monai/networks/nets/spade_network.py @virginiafdez
+/monai/networks/nets/vqvae.py @virginiafdez

.github/codecov.yml

@@ -24,8 +24,13 @@ coverage:
         flags: null
         paths: null
 
-# Disable comments on PR
-comment: false
+comment: # enable code coverage comment on PR
+  layout: "diff, flags, files"
+  behavior: default
+  require_changes: false
+  require_base: false
+  require_head: true
+  hide_project_coverage: true
 
 ignore:
   - "versioneer.py"

CONTRIBUTING.md

@@ -105,9 +105,9 @@ MONAI tests are located under `tests/`.
 A bash script (`runtests.sh`) is provided to run all tests locally.
 Please run ``./runtests.sh -h`` to see all options.
 
-To run a particular test, for example `tests/test_dice_loss.py`:
+To run a particular test, for example `tests/losses/test_dice_loss.py`:
 ```
-python -m tests.test_dice_loss
+python -m tests.losses.test_dice_loss
 ```
 
 Before submitting a pull request, we recommend that all linting and unit tests

README.md

@@ -33,6 +33,15 @@ Its ambitions are as follows:
 - customizable design for varying user expertise;
 - multi-GPU multi-node data parallelism support.
 
+## Requirements
+
+MONAI works with the [currently supported versions of Python](https://devguide.python.org/versions), and depends directly on NumPy and PyTorch with many optional dependencies.
+
+* Major releases of MONAI will have dependency versions stated for them. The current state of the `dev` branch in this repository is the unreleased development version of MONAI which typically will support current versions of dependencies and include updates and bug fixes to do so.
+* PyTorch support covers [the current version](https://github.com/pytorch/pytorch/releases) plus three previous minor versions. If compatibility issues with a PyTorch version and other dependencies arise, support for a version may be delayed until a major release.
+* Our support policy for other dependencies adheres for the most part to [SPEC0](https://scientific-python.org/specs/spec-0000), where dependency versions are supported where possible for up to two years. Discovered vulnerabilities or defects may require certain versions to be explicitly not supported.
+* See the `requirements*.txt` files for dependency version information.
+
 ## Installation
 
 To install [the current release](https://pypi.org/project/monai/), you can simply run:

docs/source/installation.md

@@ -58,10 +58,10 @@ pip install monai-weekly
 ```
 
 The weekly build is released to PyPI every Sunday with a pre-release build number `dev[%y%U]`.
-To report any issues on the weekly preview, please include the version and commit information:
+To report any issues on the weekly preview, please include the version information:
 
 ```bash
-python -c "import monai; print(monai.__version__); print(monai.__commit_id__)"
+python -c "import monai; print(monai.__version__)"
 ```
 
 Coexistence of package `monai` and `monai-weekly` in a system may cause namespace conflicts
@@ -101,20 +101,20 @@ for the latest features:
 ### Option 1 (as a part of your system-wide module):
 
 ```bash
-pip install git+https://github.com/Project-MONAI/MONAI#egg=monai
+pip install git+https://github.com/Project-MONAI/MONAI
 ```
 
 or, to build with MONAI C++/CUDA extensions:
 
 ```bash
-BUILD_MONAI=1 pip install git+https://github.com/Project-MONAI/MONAI#egg=monai
+BUILD_MONAI=1 pip install git+https://github.com/Project-MONAI/MONAI
 ```
 
 To build the extensions, if the system environment already has a version of Pytorch installed,
 `--no-build-isolation` might be preferred:
 
 ```bash
-BUILD_MONAI=1 pip install --no-build-isolation git+https://github.com/Project-MONAI/MONAI#egg=monai
+BUILD_MONAI=1 pip install --no-build-isolation git+https://github.com/Project-MONAI/MONAI
 ```
 
 this command will download and install the current `dev` branch of [MONAI from

monai/apps/auto3dseg/auto_runner.py

@@ -570,7 +570,7 @@ def set_device_info(
             self.device_setting["CUDA_VISIBLE_DEVICES"] = ",".join([str(x) for x in cuda_visible_devices])
             self.device_setting["n_devices"] = len(cuda_visible_devices)
         else:
-            logger.warn(f"Wrong format of cuda_visible_devices {cuda_visible_devices}, devices not set")
+            logger.warning(f"Wrong format of cuda_visible_devices {cuda_visible_devices}, devices not set")
 
         if num_nodes is None:
             num_nodes = int(os.environ.get("NUM_NODES", 1))

monai/bundle/workflows.py

@@ -226,7 +226,7 @@ def add_property(self, name: str, required: str, desc: str | None = None) -> Non
         if self.properties is None:
             self.properties = {}
         if name in self.properties:
-            logger.warn(f"property '{name}' already exists in the properties list, overriding it.")
+            logger.warning(f"property '{name}' already exists in the properties list, overriding it.")
         self.properties[name] = {BundleProperty.DESC: desc, BundleProperty.REQUIRED: required}
 
     def check_properties(self) -> list[str] | None:
@@ -421,7 +421,7 @@ def __init__(
             for _config_file in _config_files:
                 _config_file = Path(_config_file)
                 if _config_file.parent != config_root_path:
-                    logger.warn(
+                    logger.warning(
                         f"Not all config files are in {config_root_path}. If logging_file and meta_file are"
                         f"not specified, {config_root_path} will be used as the default config root directory."
                     )
@@ -434,11 +434,11 @@ def __init__(
         self.config_root_path = config_root_path
         logging_file = str(self.config_root_path / "logging.conf") if logging_file is None else logging_file
         if logging_file is False:
-            logger.warn(f"Logging file is set to {logging_file}, skipping logging.")
+            logger.warning(f"Logging file is set to {logging_file}, skipping logging.")
         else:
             if not os.path.isfile(logging_file):
                 if logging_file == str(self.config_root_path / "logging.conf"):
-                    logger.warn(f"Default logging file in {logging_file} does not exist, skipping logging.")
+                    logger.warning(f"Default logging file in {logging_file} does not exist, skipping logging.")
                 else:
                     raise FileNotFoundError(f"Cannot find the logging config file: {logging_file}.")
             else:
@@ -503,17 +503,17 @@ def check_properties(self) -> list[str] | None:
         """
         ret = super().check_properties()
         if self.properties is None:
-            logger.warn("No available properties had been set, skipping check.")
+            logger.warning("No available properties had been set, skipping check.")
             return None
         if ret:
-            logger.warn(f"Loaded bundle does not contain the following required properties: {ret}")
+            logger.warning(f"Loaded bundle does not contain the following required properties: {ret}")
         # also check whether the optional properties use correct ID name if existing
         wrong_props = []
         for n, p in self.properties.items():
             if not p.get(BundleProperty.REQUIRED, False) and not self._check_optional_id(name=n, property=p):
                 wrong_props.append(n)
         if wrong_props:
-            logger.warn(f"Loaded bundle defines the following optional properties with wrong ID: {wrong_props}")
+            logger.warning(f"Loaded bundle defines the following optional properties with wrong ID: {wrong_props}")
         if ret is not None:
             ret.extend(wrong_props)
         return ret

monai/inferers/inferer.py

@@ -839,6 +839,7 @@ def sample(
         mode: str = "crossattn",
         verbose: bool = True,
         seg: torch.Tensor | None = None,
+        cfg: float | None = None,
     ) -> torch.Tensor | tuple[torch.Tensor, list[torch.Tensor]]:
         """
         Args:
@@ -851,6 +852,7 @@ def sample(
             mode: Conditioning mode for the network.
             verbose: if true, prints the progression bar of the sampling process.
             seg: if diffusion model is instance of SPADEDiffusionModel, segmentation must be provided.
+            cfg: classifier-free-guidance scale, which indicates the level of strengthening on the conditioning.
         """
         if mode not in ["crossattn", "concat"]:
             raise NotImplementedError(f"{mode} condition is not supported")
@@ -877,15 +879,31 @@ def sample(
                 if isinstance(diffusion_model, SPADEDiffusionModelUNet)
                 else diffusion_model
             )
-            if mode == "concat" and conditioning is not None:
-                model_input = torch.cat([image, conditioning], dim=1)
+            if (
+                cfg is not None
+            ):  # if classifier-free guidance is used, a conditioned and unconditioned bit is generated.
+                model_input = torch.cat([image] * 2, dim=0)
+                if conditioning is not None:
+                    uncondition = torch.ones_like(conditioning)
+                    uncondition.fill_(-1)
+                    conditioning_input = torch.cat([uncondition, conditioning], dim=0)
+                else:
+                    conditioning_input = None
+            else:
+                model_input = image
+                conditioning_input = conditioning
+            if mode == "concat" and conditioning_input is not None:
+                model_input = torch.cat([model_input, conditioning_input], dim=1)
                 model_output = diffusion_model(
                     model_input, timesteps=torch.Tensor((t,)).to(input_noise.device), context=None
                 )
             else:
                 model_output = diffusion_model(
-                    image, timesteps=torch.Tensor((t,)).to(input_noise.device), context=conditioning
+                    model_input, timesteps=torch.Tensor((t,)).to(input_noise.device), context=conditioning_input
                 )
+            if cfg is not None:
+                model_output_uncond, model_output_cond = model_output.chunk(2)
+                model_output = model_output_uncond + cfg * (model_output_cond - model_output_uncond)
 
             # 2. compute previous image: x_t -> x_t-1
             if not isinstance(scheduler, RFlowScheduler):
@@ -1166,6 +1184,7 @@ def sample(  # type: ignore[override]
         mode: str = "crossattn",
         verbose: bool = True,
         seg: torch.Tensor | None = None,
+        cfg: float | None = None,
     ) -> torch.Tensor | tuple[torch.Tensor, list[torch.Tensor]]:
         """
         Args:
@@ -1180,6 +1199,7 @@ def sample(  # type: ignore[override]
             verbose: if true, prints the progression bar of the sampling process.
             seg: if diffusion model is instance of SPADEDiffusionModel, or autoencoder_model
              is instance of SPADEAutoencoderKL, segmentation must be provided.
+            cfg: classifier-free-guidance scale, which indicates the level of strengthening on the conditioning.
         """
 
         if (
@@ -1203,6 +1223,7 @@ def sample(  # type: ignore[override]
             mode=mode,
             verbose=verbose,
             seg=seg,
+            cfg=cfg,
         )
 
         if save_intermediates:
@@ -1381,6 +1402,7 @@ def sample(  # type: ignore[override]
         mode: str = "crossattn",
         verbose: bool = True,
         seg: torch.Tensor | None = None,
+        cfg: float | None = None,
     ) -> torch.Tensor | tuple[torch.Tensor, list[torch.Tensor]]:
         """
         Args:
@@ -1395,6 +1417,7 @@ def sample(  # type: ignore[override]
             mode: Conditioning mode for the network.
             verbose: if true, prints the progression bar of the sampling process.
             seg: if diffusion model is instance of SPADEDiffusionModel, segmentation must be provided.
+                        cfg: classifier-free-guidance scale, which indicates the level of strengthening on the conditioning.
         """
         if mode not in ["crossattn", "concat"]:
             raise NotImplementedError(f"{mode} condition is not supported")
@@ -1413,14 +1436,31 @@ def sample(  # type: ignore[override]
             progress_bar = iter(zip(scheduler.timesteps, all_next_timesteps))
         intermediates = []
 
+        if cfg is not None:
+            cn_cond = torch.cat([cn_cond] * 2, dim=0)
+
         for t, next_t in progress_bar:
+            # Controlnet prediction
+            if cfg is not None:
+                model_input = torch.cat([image] * 2, dim=0)
+                if conditioning is not None:
+                    uncondition = torch.ones_like(conditioning)
+                    uncondition.fill_(-1)
+                    conditioning_input = torch.cat([uncondition, conditioning], dim=0)
+                else:
+                    conditioning_input = None
+            else:
+                model_input = image
+                conditioning_input = conditioning
+
+            # Diffusion model prediction
             diffuse = diffusion_model
             if isinstance(diffusion_model, SPADEDiffusionModelUNet):
                 diffuse = partial(diffusion_model, seg=seg)
 
-            if mode == "concat" and conditioning is not None:
+            if mode == "concat" and conditioning_input is not None:
                 # 1. Conditioning
-                model_input = torch.cat([image, conditioning], dim=1)
+                model_input = torch.cat([model_input, conditioning_input], dim=1)
                 # 2. ControlNet forward
                 down_block_res_samples, mid_block_res_sample = controlnet(
                     x=model_input,
@@ -1437,20 +1477,28 @@ def sample(  # type: ignore[override]
                     mid_block_additional_residual=mid_block_res_sample,
                 )
             else:
+                # 1. Controlnet forward
                 down_block_res_samples, mid_block_res_sample = controlnet(
-                    x=image,
+                    x=model_input,
                     timesteps=torch.Tensor((t,)).to(input_noise.device),
                     controlnet_cond=cn_cond,
-                    context=conditioning,
+                    context=conditioning_input,
                 )
+                # 2. predict noise model_output
                 model_output = diffuse(
-                    image,
+                    model_input,
                     timesteps=torch.Tensor((t,)).to(input_noise.device),
-                    context=conditioning,
+                    context=conditioning_input,
                     down_block_additional_residuals=down_block_res_samples,
                     mid_block_additional_residual=mid_block_res_sample,
                 )
 
+            # If classifier-free guidance isn't None, we split and compute the weighting between
+            # conditioned and unconditioned output.
+            if cfg is not None:
+                model_output_uncond, model_output_cond = model_output.chunk(2)
+                model_output = model_output_uncond + cfg * (model_output_cond - model_output_uncond)
+
             # 3. compute previous image: x_t -> x_t-1
             if not isinstance(scheduler, RFlowScheduler):
                 image, _ = scheduler.step(model_output, t, image)  # type: ignore
@@ -1714,6 +1762,7 @@ def sample(  # type: ignore[override]
         mode: str = "crossattn",
         verbose: bool = True,
         seg: torch.Tensor | None = None,
+        cfg: float | None = None,
     ) -> torch.Tensor | tuple[torch.Tensor, list[torch.Tensor]]:
         """
         Args:
@@ -1730,6 +1779,7 @@ def sample(  # type: ignore[override]
             verbose: if true, prints the progression bar of the sampling process.
             seg: if diffusion model is instance of SPADEDiffusionModel, or autoencoder_model
              is instance of SPADEAutoencoderKL, segmentation must be provided.
+            cfg: classifier-free-guidance scale, which indicates the level of strengthening on the conditioning.
         """
 
         if (
@@ -1757,6 +1807,7 @@ def sample(  # type: ignore[override]
             mode=mode,
             verbose=verbose,
             seg=seg,
+            cfg=cfg,
         )
 
         if save_intermediates:

tests/bundle/test_bundle_download.py

@@ -316,20 +316,6 @@ def test_load_weights(self, bundle_files, bundle_name, repo, device, model_file)
                 output_2 = model_2.forward(input_tensor)
                 assert_allclose(output_2, expected_output, atol=1e-4, rtol=1e-4, type_test=False)
 
-                model_3 = load(
-                    name=bundle_name,
-                    model_file=model_file,
-                    bundle_dir=tempdir,
-                    progress=False,
-                    device=device,
-                    net_name=model_name,
-                    source="github",
-                    **net_args,
-                )
-                model_3.eval()
-                output_3 = model_3.forward(input_tensor)
-                assert_allclose(output_3, expected_output, atol=1e-4, rtol=1e-4, type_test=False)
-
     @parameterized.expand([TEST_CASE_8])
     @skip_if_quick
     @skipUnless(has_huggingface_hub, "Requires `huggingface_hub`.")

tests/inferers/test_controlnet_inferers.py

@@ -482,16 +482,20 @@ def test_sample_intermediates(self, model_params, controlnet_params, input_shape
         scheduler = DDPMScheduler(num_train_timesteps=10)
         inferer = ControlNetDiffusionInferer(scheduler=scheduler)
         scheduler.set_timesteps(num_inference_steps=10)
-        sample, intermediates = inferer.sample(
-            input_noise=noise,
-            diffusion_model=model,
-            scheduler=scheduler,
-            controlnet=controlnet,
-            cn_cond=mask,
-            save_intermediates=True,
-            intermediate_steps=1,
-        )
-        self.assertEqual(len(intermediates), 10)
+
+        for cfg in [5, None]:
+            sample, intermediates = inferer.sample(
+                input_noise=noise,
+                diffusion_model=model,
+                scheduler=scheduler,
+                controlnet=controlnet,
+                cn_cond=mask,
+                save_intermediates=True,
+                intermediate_steps=1,
+                cfg=cfg,
+            )
+
+            self.assertEqual(len(intermediates), 10)
 
     @parameterized.expand(CNDM_TEST_CASES)
     @skipUnless(has_einops, "Requires einops")