Merge branch 'Project-MONAI:main' into main

Project-MONAI · Sep 16, 2024 · c01dacb · c01dacb
2 parents e5fba2b + ae3f383
commit c01dacb
Show file tree

Hide file tree

Showing 18 changed files with 38 additions and 34 deletions.
diff --git a/3d_segmentation/unetr_btcv_segmentation_3d.ipynb b/3d_segmentation/unetr_btcv_segmentation_3d.ipynb
@@ -19,7 +19,6 @@
     "\n",
     "\n",
     "This tutorial demonstrates how to construct a training workflow of UNETR [1] on multi-organ segmentation task using the BTCV challenge dataset.\n",
-    "![image](https://lh3.googleusercontent.com/pw/AM-JKLU2eTW17rYtCmiZP3WWC-U1HCPOHwLe6pxOfJXwv2W-00aHfsNy7jeGV1dwUq0PXFOtkqasQ2Vyhcu6xkKsPzy3wx7O6yGOTJ7ZzA01S6LSh8szbjNLfpbuGgMe6ClpiS61KGvqu71xXFnNcyvJNFjN=w1448-h496-no?authuser=0)\n",
     "\n",
     "And it contains the following features:\n",
     "1. Transforms for dictionary format data.\n",
@@ -51,7 +50,7 @@
     "3. [Efficient multi-atlas abdominal segmentation on clinically acquired CT with SIMPLE context learning (MIA)](https://www.sciencedirect.com/science/article/abs/pii/S1361841515000766?via%3Dihub)\n",
     "\n",
     "\n",
-    "![image](https://lh3.googleusercontent.com/pw/AM-JKLX0svvlMdcrchGAgiWWNkg40lgXYjSHsAAuRc5Frakmz2pWzSzf87JQCRgYpqFR0qAjJWPzMQLc_mmvzNjfF9QWl_1OHZ8j4c9qrbR6zQaDJWaCLArRFh0uPvk97qAa11HtYbD6HpJ-wwTCUsaPcYvM=w1724-h522-no?authuser=0)\n",
+    "![image](../figures/BTCV_organs.png)\n",
     "\n",
     "\n",
     "\n",
@@ -586,7 +585,7 @@
     "    hidden_size=768,\n",
     "    mlp_dim=3072,\n",
     "    num_heads=12,\n",
-    "    pos_embed=\"perceptron\",\n",
+    "    proj_type=\"perceptron\",\n",
     "    norm_name=\"instance\",\n",
     "    res_block=True,\n",
     "    dropout_rate=0.0,\n",

diff --git a/3d_segmentation/unetr_btcv_segmentation_3d_lightning.ipynb b/3d_segmentation/unetr_btcv_segmentation_3d_lightning.ipynb
@@ -423,7 +423,7 @@
     "            hidden_size=768,\n",
     "            mlp_dim=3072,\n",
     "            num_heads=12,\n",
-    "            pos_embed=\"perceptron\",\n",
+    "            proj_type=\"perceptron\",\n",
     "            norm_name=\"instance\",\n",
     "            res_block=True,\n",
     "            conv_block=True,\n",

diff --git a/deepedit/ignite/train.py b/deepedit/ignite/train.py
@@ -76,7 +76,7 @@ def get_network(network, labels, spatial_size):
             hidden_size=1536,
             mlp_dim=3072,
             num_heads=48,
-            pos_embed="conv",
+            proj_type="conv",
             norm_name="instance",
             res_block=True,
         )

diff --git a/deployment/Triton/client/client.py b/deployment/Triton/client/client.py
@@ -51,7 +51,7 @@
 from monai.apps.utils import download_and_extract
 
 model_name = "monai_covid"
-gdrive_path = "https://drive.google.com/uc?id=1GYvHGU2jES0m_msin-FFQnmTOaHkl0LN"
+gdrive_path = "https://developer.download.nvidia.com/assets/Clara/monai/tutorials/covid19_compressed.tar.gz"
 covid19_filename = "covid19_compress.tar.gz"
 md5_check = "cadd79d5ca9ccdee2b49cd0c8a3e6217"
 

diff --git a/deployment/Triton/client/client_mednist.py b/deployment/Triton/client/client_mednist.py
@@ -55,7 +55,7 @@
 
 
 model_name = "mednist_class"
-gdrive_path = "https://drive.google.com/uc?id=1HQk4i4vXKUX_aAYR4wcZQKd-qk5Lcm_W"
+gdrive_path = "https://developer.download.nvidia.com/assets/Clara/monai/tutorials/MedNIST_demo.tar.gz"
 mednist_filename = "MedNIST_demo.tar.gz"
 md5_check = "3f24a5833bb0455a7815c4e0ecc8a810"
 

diff --git a/deployment/Triton/models/mednist_class/1/model.py b/deployment/Triton/models/mednist_class/1/model.py
@@ -74,7 +74,7 @@
 
 
 logger = logging.getLogger(__name__)
-gdrive_url = "https://drive.google.com/uc?id=1c6noLV9oR0_mQwrsiQ9TqaaeWFKyw46l"
+gdrive_url = "https://developer.download.nvidia.com/assets/Clara/monai/tutorials/MedNist_model.tar.gz"
 model_filename = "MedNIST_model.tar.gz"
 md5_check = "a4fb9d6147599e104b5d8dc1809ed034"
 

diff --git a/deployment/Triton/models/monai_covid/1/model.py b/deployment/Triton/models/monai_covid/1/model.py
@@ -64,7 +64,7 @@
 
 
 logger = logging.getLogger(__name__)
-gdrive_url = "https://drive.google.com/uc?id=1U9Oaw47SWMJeDkg1FSTY1W__tQOY1nAZ"
+gdrive_url = "https://developer.download.nvidia.com/assets/Clara/monai/tutorials/covid19_model.tar.gz"
 model_filename = "covid19_model.tar.gz"
 md5_check = "571046a25659515bf7abee4266f14435"
 

diff --git a/detection/generate_transforms.py b/detection/generate_transforms.py
@@ -148,7 +148,7 @@ def generate_detection_train_transform(
                 spatial_axes=(0, 1),
             ),
             # apply the same affine matrix which already applied on the images to the points
-            ApplyTransformToPointsd(keys=[point_key], refer_key=image_key, affine_lps_to_ras=affine_lps_to_ras),
+            ApplyTransformToPointsd(keys=[point_key], refer_keys=image_key, affine_lps_to_ras=affine_lps_to_ras),
             # convert points back to boxes
             ConvertPointsToBoxesd(keys=[point_key]),
             ClipBoxToImaged(

diff --git a/generation/2d_ldm/README.md b/generation/2d_ldm/README.md
@@ -26,12 +26,9 @@ python download_brats_data.py -e ./config/environment.json
 
 Disclaimer: We are not the host of the data. Please make sure to read the requirements and usage policies of the data and give credit to the authors of the dataset!
 
-### 2. Installation
-Please refer to the [Installation of MONAI Generative Model](../README.md)
+### 2. Run the example
 
-### 3. Run the example
-
-#### [3.1 2D Autoencoder Training](./train_autoencoder.py)
+#### [2.1 2D Autoencoder Training](./train_autoencoder.py)
 
 The network configuration files are located in [./config/config_train_32g.json](./config/config_train_32g.json) for 32G GPU and [./config/config_train_16g.json](./config/config_train_16g.json) for 16G GPU. You can modify the hyperparameters in these files to suit your requirements.
 
@@ -74,7 +71,7 @@ An example reconstruction result is shown below:
   <img src="./figs/recon.png" alt="Autoencoder reconstruction result")
 </p>
 
-#### [3.2 2D Latent Diffusion Training](./train_diffusion.py)
+#### [2.2 2D Latent Diffusion Training](./train_diffusion.py)
 The training script uses the batch size and patch size defined in the configuration files. If you have a different GPU memory size, you should adjust the `"batch_size"` and `"patch_size"` parameters in the `"diffusion_train"` to match your GPU. Note that the `"patch_size"` needs to be divisible by 16 and no larger than 256.
 
 To train with single 32G GPU, please run:
@@ -97,7 +94,7 @@ torchrun \
   <img src="./figs/val_diffusion.png" alt="latent diffusion validation curve" width="45%" >
 </p>
 
-#### [3.3 Inference](./inference.py)
+#### [2.3 Inference](./inference.py)
 To generate one image during inference, please run the following command:
 ```bash
 python inference.py -c ./config/config_train_32g.json -e ./config/environment.json --num 1
@@ -115,7 +112,7 @@ An example output is shown below.
   <img src="./figs/syn_3.jpeg" width="20%" >
 </p>
 
-### 4. Questions and bugs
+### 3. Questions and bugs
 
 - For questions relating to the use of MONAI, please use our [Discussions tab](https://github.com/Project-MONAI/MONAI/discussions) on the main repository of MONAI.
 - For bugs relating to MONAI functionality, please create an issue on the [main repository](https://github.com/Project-MONAI/MONAI/issues).

diff --git a/generation/3d_ldm/README.md b/generation/3d_ldm/README.md
@@ -26,12 +26,9 @@ python download_brats_data.py -e ./config/environment.json
 
 Disclaimer: We are not the host of the data. Please make sure to read the requirements and usage policies of the data and give credit to the authors of the dataset!
 
-### 2. Installation
-Please refer to the [Installation of MONAI Generative Model](../README.md)
+### 2. Run the example
 
-### 3. Run the example
-
-#### [3.1 3D Autoencoder Training](./train_autoencoder.py)
+#### [2.1 3D Autoencoder Training](./train_autoencoder.py)
 
 The network configuration files are located in [./config/config_train_32g.json](./config/config_train_32g.json) for 32G GPU
 and [./config/config_train_16g.json](./config/config_train_16g.json) for 16G GPU.
@@ -73,7 +70,7 @@ torchrun \
 
 With eight DGX1V 32G GPUs, it took around 55 hours to train 1000 epochs.
 
-#### [3.2 3D Latent Diffusion Training](./train_diffusion.py)
+#### [2.2 3D Latent Diffusion Training](./train_diffusion.py)
 The training script uses the batch size and patch size defined in the configuration files. If you have a different GPU memory size, you should adjust the `"batch_size"` and `"patch_size"` parameters in the `"diffusion_train"` to match your GPU. Note that the `"patch_size"` needs to be divisible by 16.
 
 To train with single 32G GPU, please run:
@@ -96,7 +93,7 @@ torchrun \
   <img src="./figs/val_diffusion.png" alt="latent diffusion validation curve" width="45%" >
 </p>
 
-#### [3.3 Inference](./inference.py)
+#### [2.3 Inference](./inference.py)
 To generate one image during inference, please run the following command:
 ```bash
 python inference.py -c ./config/config_train_32g.json -e ./config/environment.json --num 1
@@ -112,7 +109,7 @@ An example output is shown below.
   <img src="./figs/syn_cor.png" width="30%" >
 </p>
 
-### 4. Questions and bugs
+### 3. Questions and bugs
 
 - For questions relating to the use of MONAI, please use our [Discussions tab](https://github.com/Project-MONAI/MONAI/discussions) on the main repository of MONAI.
 - For bugs relating to MONAI functionality, please create an issue on the [main repository](https://github.com/Project-MONAI/MONAI/issues).

diff --git a/generation/maisi/README.md b/generation/maisi/README.md
@@ -68,6 +68,14 @@ The information for the inference input, like body region and anatomy to generat
 
 To generate images with substantial dimensions, such as 512 &times; 512 &times; 512 or larger, using GPUs with 80GB of memory, it is advisable to configure the `"num_splits"` parameter in [the auto-encoder configuration](./configs/config_maisi.json#L11-L37) to 16. This adjustment is crucial to avoid out-of-memory issues during inference.
 
+#### Recommended spacing for different output sizes:
+
+|`output_size`| Recommended `"spacing"`|
+|:-----:|:-----:|
+[256, 256, 256]  | [1.5, 1.5, 1.5] |
+[512, 512, 128]  | [0.8, 0.8, 2.5] |
+[512, 512, 512]  | [1.0, 1.0, 1.0] |
+
 #### Execute Inference:
 To run the inference script, please run:
 ```bash

diff --git a/generation/maisi/scripts/infer_controlnet.py b/generation/maisi/scripts/infer_controlnet.py
@@ -23,7 +23,7 @@
 from monai.transforms import SaveImage
 from monai.utils import RankFilter
 
-from .sample import ldm_conditional_sample_one_image
+from .sample import check_input, ldm_conditional_sample_one_image
 from .utils import define_instance, prepare_maisi_controlnet_json_dataloader, setup_ddp
 
 
@@ -150,10 +150,13 @@ def main():
         top_region_index_tensor = batch["top_region_index"].to(device)
         bottom_region_index_tensor = batch["bottom_region_index"].to(device)
         spacing_tensor = batch["spacing"].to(device)
+        out_spacing = tuple((batch["spacing"].squeeze().numpy() / 100).tolist())
         # get target dimension
         dim = batch["dim"]
         output_size = (dim[0].item(), dim[1].item(), dim[2].item())
         latent_shape = (args.latent_channels, output_size[0] // 4, output_size[1] // 4, output_size[2] // 4)
+        # check if output_size and out_spacing are valid.
+        check_input(None, None, None, output_size, out_spacing, None)
         # generate a single synthetic image using a latent diffusion model with controlnet.
         synthetic_images, _ = ldm_conditional_sample_one_image(
             autoencoder,

diff --git a/modules/developer_guide.ipynb b/modules/developer_guide.ipynb
@@ -717,7 +717,7 @@
     "id": "kvn_6mf9gZoA"
    },
    "source": [
-    "The following commands will start a `SupervisedTrainer` instance. As an extension of Pytorch ignite's facilities, it combines all the elements mentioned before. Calling `trainer.run()` will train the network for two epochs and compute `MeadDice` metric based on the training data at the end of every epoch.\n",
+    "The following commands will start a `SupervisedTrainer` instance. As an extension of Pytorch ignite's facilities, it combines all the elements mentioned before. Calling `trainer.run()` will train the network for two epochs and compute `MeanDice` metric based on the training data at the end of every epoch.\n",
     "\n",
     "The `key_train_metric` is used to track the progress of model quality improvement. Additional handlers could be set to do early stopping and learning rate scheduling.\n",
     "\n",

diff --git a/modules/interpretability/class_lung_lesion.ipynb b/modules/interpretability/class_lung_lesion.ipynb
@@ -29,7 +29,7 @@
     "\n",
     "For the demo data:\n",
     "- Please see the `bbox_gen.py` script for generating the patch classification data from MSD task06_lung (available via `monai.apps.DecathlonDataset`);\n",
-    "- Alternatively, the patch dataset (~130MB) is available for direct downloading at: https://drive.google.com/drive/folders/1pQdzdkkC9c2GOblLgpGlG3vxsSK9NtDx\n",
+    "- Alternatively, the patch dataset (~130MB) is available for direct downloading at: https://github.com/Project-MONAI/MONAI-extra-test-data/releases/download/0.8.1/lung_lesion_patches.tar.gz\n",
     "\n",
     "[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/Project-MONAI/tutorials/blob/main/modules/interpretability/class_lung_lesion.ipynb)"
    ]

diff --git a/modules/tcia_csv_processing.ipynb b/modules/tcia_csv_processing.ipynb
@@ -247,8 +247,8 @@
    "metadata": {},
    "source": [
     "## Download and Load the CSV file with `TCIADataset`\n",
-    "Here we use the demo data in Google drive:\n",
-    "https://drive.google.com/file/d/1HQ7BZvBr1edmi8HIwdG5KBweXWms5Uzk/view?usp=sharing \n",
+    "Here we use the demo data located here:\n",
+    "https://github.com/Project-MONAI/MONAI-extra-test-data/releases/download/0.8.1/ISPY1_Combined.csv \n",
     "\n",
     "Expect the first row of CSV file to be titles of columns. we only use the first 8 rows to execute demo processing."
    ]

diff --git a/self_supervised_pretraining/vit_unetr_ssl/multi_gpu/mgpu_ssl_train.py b/self_supervised_pretraining/vit_unetr_ssl/multi_gpu/mgpu_ssl_train.py
@@ -175,7 +175,7 @@ def main(args):
         in_channels=1,
         img_size=(96, 96, 96),
         patch_size=(16, 16, 16),
-        pos_embed="conv",
+        proj_type="conv",
         hidden_size=768,
         mlp_dim=3072,
     )

diff --git a/self_supervised_pretraining/vit_unetr_ssl/ssl_finetune.ipynb b/self_supervised_pretraining/vit_unetr_ssl/ssl_finetune.ipynb
@@ -317,7 +317,7 @@
     "    hidden_size=768,\n",
     "    mlp_dim=3072,\n",
     "    num_heads=12,\n",
-    "    pos_embed=\"conv\",\n",
+    "    proj_type=\"conv\",\n",
     "    norm_name=\"instance\",\n",
     "    res_block=True,\n",
     "    dropout_rate=0.0,\n",

diff --git a/self_supervised_pretraining/vit_unetr_ssl/ssl_train.ipynb b/self_supervised_pretraining/vit_unetr_ssl/ssl_train.ipynb
@@ -253,7 +253,7 @@
     "    in_channels=1,\n",
     "    img_size=(96, 96, 96),\n",
     "    patch_size=(16, 16, 16),\n",
-    "    pos_embed=\"conv\",\n",
+    "    proj_type=\"conv\",\n",
     "    hidden_size=768,\n",
     "    mlp_dim=3072,\n",
     ")\n",