Update to scripts

README.md

@@ -46,11 +46,11 @@ The dataset used for training unfortunately cannot be made public, however the t
 * 200: Tricuspid septal
 * 250: Tricuspid free wall
 
-The following command will train with the default NPZ filename `./valvelandmarks.npz`:
+The following command will train with the default NPZ filename `./valvelandmarks.npz`, assuming the current directory is the bundle directory:
 
 ```sh
-PYTHONPATH=./scripts python -m monai.bundle run training --meta_file configs/metadata.json \
-    --config_file configs/train.json --bundle_root . --dataset_file /path/to/data --output_dir /path/to/outputs
+python -m monai.bundle run training --meta_file configs/metadata.json --config_file "['configs/train.json', 'configs/common.json']" \
+    --bundle_root . --dataset_file ./valvelandmarks.npz --output_dir /path/to/outputs
 ```
 
 ## Inference
@@ -58,11 +58,11 @@ PYTHONPATH=./scripts python -m monai.bundle run training --meta_file configs/met
 The included `inference.json` script will run inference on a directory containing Nifti files whose images have shape `(256, 256, 1, N)` for `N` timesteps. For each image the output in the `output_dir` directory will be a npy file containing a result array of shape `(N, 2, 10)` storing the 10 coordinates for each `N` timesteps. Invoking this script can be done as follows, assuming the current directory is the bundle directory:
 
 ```sh
-PYTHONPATH=./scripts python -m monai.bundle run evaluating --meta_file configs/metadata.json \
-    --config_file configs/inference.json --bundle_root . --dataset_dir /path/to/data --output_dir /path/to/outputs
+python -m monai.bundle run evaluating --meta_file configs/metadata.json --config_file "['configs/inference.json', 'configs/common.json']" \
+    --bundle_root . --dataset_dir /path/to/data --output_dir /path/to/outputs
 ```
 
-It is important to set the `PYTHONPATH` variable since code in the provided scripts directory is necessary for inference. The provided test Nifti file can be placed in a directory which is then used as the `dataset_dir` value. This image was derived from [the AMRG Cardiac Atlas dataset](http://www.cardiacatlas.org/studies/amrg-cardiac-atlas) (AMRG Cardiac Atlas, Auckland MRI Research Group, Auckland, New Zealand). The results from this inference can be visualised by changing path values in [view_results.ipynb](./view_results.ipynb).
+The provided test Nifti file can be placed in a directory which is then used as the `dataset_dir` value. This image was derived from [the AMRG Cardiac Atlas dataset](http://www.cardiacatlas.org/studies/amrg-cardiac-atlas) (AMRG Cardiac Atlas, Auckland MRI Research Group, Auckland, New Zealand). The results from this inference can be visualised by changing path values in [view_results.ipynb](./view_results.ipynb).
 
 
 ### Reference

configs/inference.json

@@ -2,11 +2,12 @@
     "imports": [
         "$import os",
         "$import glob",
+        "$import torch",
         "$import scripts"
     ],
     "bundle_root": ".",
-    "device": "$torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')",
     "ckpt_path": "$@bundle_root + '/models/model.pt'",
+    "device": "$torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')",
     "dataset_dir": "/workspace/data",
     "datalist": "$list(sorted(glob.glob(@dataset_dir + '/*.nii*')))",
     "output_dir": "./output",

configs/metadata.json

@@ -1,11 +1,12 @@
 {
     "schema": "https://github.com/Project-MONAI/MONAI-extra-test-data/releases/download/0.8.1/meta_schema_20220729.json",
-    "version": "0.2.0",
+    "version": "0.3.0",
     "changelog": {
-        "0.2.0": "unify naming",
+        "0.3.0": "Update to scripts",
+        "0.2.0": "Unify naming",
         "0.1.0": "Initial version"
     },
-    "monai_version": "1.0.0rc1",
+    "monai_version": "1.0.0",
     "pytorch_version": "1.10.2",
     "numpy_version": "1.21.2",
     "optional_packages_version": {},

configs/train.json

@@ -1,15 +1,21 @@
 {
     "imports": [
         "$import datetime",
-        "$import numpy as np",
+        "$import numpy",
         "$import torch",
         "$import ignite",
         "$import scripts"
     ],
     "bundle_root": ".",
-    "val_interval": 1,
-    "device": "$torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')",
     "ckpt_path": "$@bundle_root + '/models/model.pt'",
+    "device": "$torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')",
+    "val_interval": 1,
+    "num_iters": 400,
+    "batch_size": 600,
+    "num_epochs": 100,
+    "num_substeps": 3,
+    "learning_rate": 0.0001,
+    "num_workers": 8,
     "dataset_file": "./valvelandmarks.npz",
     "output_dir": "$datetime.datetime.now().strftime('./results/output_%y%m%d_%H%M%S')",
     "network_def": {
@@ -56,7 +62,7 @@
                 "_target_": "EnsureTyped",
                 "keys": "@both_keys",
                 "data_type": "numpy",
-                "dtype": "$(np.float32, np.int32)"
+                "dtype": "$(numpy.float32, numpy.int32)"
             },
             {
                 "_target_": "EnsureTyped",
@@ -147,7 +153,7 @@
                 "_target_": "EnsureTyped",
                 "keys": "@both_keys",
                 "data_type": "numpy",
-                "dtype": "$(np.float32, np.int32)"
+                "dtype": "$(numpy.float32, numpy.int32)"
             },
             {
                 "_target_": "EnsureTyped",
@@ -186,10 +192,6 @@
         },
         "transform": "@eval_transforms"
     },
-    "num_iters": 400,
-    "batch_size": 600,
-    "num_epochs": 100,
-    "num_substeps": 3,
     "sampler": {
         "_target_": "torch.utils.data.WeightedRandomSampler",
         "weights": "$torch.ones(len(@train_dataset))",
@@ -201,14 +203,14 @@
         "dataset": "@train_dataset",
         "batch_size": "@batch_size",
         "repeats": "@num_substeps",
-        "num_workers": 8,
+        "num_workers": "@num_workers",
         "sampler": "@sampler"
     },
     "eval_dataloader": {
         "_target_": "DataLoader",
         "dataset": "@eval_dataset",
         "batch_size": "@batch_size",
-        "num_workers": 8
+        "num_workers": "@num_workers"
     },
     "lossfn": {
         "_target_": "torch.nn.L1Loss"
@@ -216,7 +218,7 @@
     "optimizer": {
         "_target_": "torch.optim.Adam",
         "params": "[email protected]()",
-        "lr": 0.0001
+        "lr": "@learning_rate"
     },
     "evaluator": {
         "_target_": "SupervisedEvaluator",

docs/README.md

@@ -39,11 +39,11 @@ The dataset used for training unfortunately cannot be made public, however the t
 * 200: Tricuspid septal
 * 250: Tricuspid free wall
 
-The following command will train with the default NPZ filename `./valvelandmarks.npz`:
+The following command will train with the default NPZ filename `./valvelandmarks.npz`, assuming the current directory is the bundle directory:
 
 ```sh
-PYTHONPATH=./scripts python -m monai.bundle run training --meta_file configs/metadata.json \
-    --config_file configs/train.json --bundle_root . --dataset_file /path/to/data --output_dir /path/to/outputs
+python -m monai.bundle run training --meta_file configs/metadata.json --config_file "['configs/train.json', 'configs/common.json']" \
+    --bundle_root . --dataset_file ./valvelandmarks.npz --output_dir /path/to/outputs
 ```
 
 ## Inference
@@ -51,11 +51,11 @@ PYTHONPATH=./scripts python -m monai.bundle run training --meta_file configs/met
 The included `inference.json` script will run inference on a directory containing Nifti files whose images have shape `(256, 256, 1, N)` for `N` timesteps. For each image the output in the `output_dir` directory will be a npy file containing a result array of shape `(N, 2, 10)` storing the 10 coordinates for each `N` timesteps. Invoking this script can be done as follows, assuming the current directory is the bundle directory:
 
 ```sh
-PYTHONPATH=./scripts python -m monai.bundle run evaluating --meta_file configs/metadata.json \
-    --config_file configs/inference.json --bundle_root . --dataset_dir /path/to/data --output_dir /path/to/outputs
+python -m monai.bundle run evaluating --meta_file configs/metadata.json --config_file "['configs/inference.json', 'configs/common.json']" \
+    --bundle_root . --dataset_dir /path/to/data --output_dir /path/to/outputs
 ```
 
-It is important to set the `PYTHONPATH` variable since code in the provided scripts directory is necessary for inference. The provided test Nifti file can be placed in a directory which is then used as the `dataset_dir` value. This image was derived from [the AMRG Cardiac Atlas dataset](http://www.cardiacatlas.org/studies/amrg-cardiac-atlas) (AMRG Cardiac Atlas, Auckland MRI Research Group, Auckland, New Zealand). The results from this inference can be visualised by changing path values in [view_results.ipynb](./view_results.ipynb).
+The provided test Nifti file can be placed in a directory which is then used as the `dataset_dir` value. This image was derived from [the AMRG Cardiac Atlas dataset](http://www.cardiacatlas.org/studies/amrg-cardiac-atlas) (AMRG Cardiac Atlas, Auckland MRI Research Group, Auckland, New Zealand). The results from this inference can be visualised by changing path values in [view_results.ipynb](./view_results.ipynb).
 
 
 ### Reference

scripts/__init__.py

@@ -0,0 +1 @@
+from . import valve_landmarks