example tasks: organize imports and ignore E402 for ncalab imports

Author: henrykrumb

tasks/class_cifar10/README.md

@@ -0,0 +1,3 @@
+# CIFAR10 Classification
+
+We currently achieve about 75% test set accuracy with NCAs.

tasks/class_cifar10/eval_class_cifar10.py

@@ -10,10 +10,9 @@
 from torchvision import transforms  # type: ignore[import-untyped]
 from torchvision.transforms import v2  # type: ignore[import-untyped]
 from tqdm import tqdm
-
 from train_class_cifar10 import (
-    WEIGHTS_PATH,
     TASK_PATH,
+    WEIGHTS_PATH,
     alive_mask,
     default_hidden_channels,
     fire_rate,
@@ -25,7 +24,11 @@
 sys.path.append(root_dir)
 
 
-from ncalab import ClassificationNCAModel, get_compute_device, fix_random_seed
+from ncalab import (  # noqa: E402
+    ClassificationNCAModel,
+    fix_random_seed,
+    get_compute_device,
+)
 
 T = transforms.Compose(
     [

tasks/class_cifar10/train_class_cifar10.py

@@ -6,15 +6,15 @@
 import click
 import numpy as np  # type: ignore[import-untyped]
 import torch  # type: ignore[import-untyped]
+import torchvision  # type: ignore[import-untyped]
 from torch.utils.data.sampler import SubsetRandomSampler  # type: ignore[import-untyped]
 from torch.utils.tensorboard import SummaryWriter  # type: ignore[import-untyped]
-import torchvision  # type: ignore[import-untyped]
 from torchvision import transforms  # type: ignore[import-untyped]
 from torchvision.transforms import v2  # type: ignore[import-untyped]
 
 root_dir = os.path.abspath(os.path.join(os.path.dirname(__file__), "../.."))
 sys.path.append(root_dir)
-from ncalab import (
+from ncalab import (  # noqa: E402
     BasicNCATrainer,
     ClassificationNCAModel,
     get_compute_device,
@@ -61,18 +61,20 @@ def train_class_cifar10(
             transforms.RandomPerspective(distortion_scale=0.5, p=0.5),
             transforms.RandomHorizontalFlip(),
             transforms.RandomVerticalFlip(),
-            transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5)),
+            transforms.RandomCrop(32, padding=4),
+            transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010)),
         ]
     )
     T_val = transforms.Compose(
         [
             v2.ToImage(),
             v2.ToDtype(torch.float, scale=True),
             v2.ConvertImageDtype(dtype=torch.float32),
-            transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5)),
+            transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010)),
         ]
     )
 
+    # Split train dataset into train and validation
     train_dataset = torchvision.datasets.CIFAR10(
         root=TASK_PATH / "data",
         train=True,
@@ -86,7 +88,6 @@ def train_class_cifar10(
         download=True,
         transform=T_val,
     )
-
     indices = list(range(len(train_dataset)))
     split = int(np.floor(0.1 * len(train_dataset)))
     np.random.shuffle(indices)
@@ -122,8 +123,9 @@ def train_class_cifar10(
         "ship",
         "truck",
     ]
-
     num_classes = len(class_names)
+
+    # Create NCA model for classification
     nca = ClassificationNCAModel(
         device,
         num_image_channels=3,
@@ -135,11 +137,12 @@ def train_class_cifar10(
         use_temporal_encoding=use_temporal_encoding,
         class_names=class_names,
     )
+    # Train the NCA model
     trainer = BasicNCATrainer(
         nca,
         WEIGHTS_PATH / "classification_cifar10",
         batch_repeat=2,
-        max_epochs=100,
+        max_epochs=500,
         gradient_clipping=gradient_clipping,
         steps_range=(32, 48),
         steps_validation=42,

tasks/class_medmnist/eval_class_bloodmnist.py

@@ -1,34 +1,30 @@
 #!/usr/bin/env python3
-import sys
 import os
-
-root_dir = os.path.abspath(os.path.join(os.path.dirname(__file__), "../.."))
-sys.path.append(root_dir)
-
-from ncalab import ClassificationNCAModel, get_compute_device
+import sys
 
 import click
-
-from medmnist import INFO, BloodMNIST  # type: ignore[import-untyped]
-
 import torch  # type: ignore[import-untyped]
-from torchvision import transforms  # type: ignore[import-untyped]
-from torchvision.transforms import v2  # type: ignore[import-untyped]
-
 import torchmetrics
 import torchmetrics.classification
-
+from medmnist import INFO, BloodMNIST  # type: ignore[import-untyped]
+from torchvision import transforms  # type: ignore[import-untyped]
+from torchvision.transforms import v2  # type: ignore[import-untyped]
 from tqdm import tqdm
-
 from train_class_bloodmnist import (
-    pad_noise,
-    alive_mask,
-    use_temporal_encoding,
-    fire_rate,
     WEIGHTS_PATH,
+    alive_mask,
     default_hidden_channels,
+    fire_rate,
+    pad_noise,
+    use_temporal_encoding,
 )
 
+root_dir = os.path.abspath(os.path.join(os.path.dirname(__file__), "../.."))
+sys.path.append(root_dir)
+
+
+from ncalab import ClassificationNCAModel, get_compute_device  # noqa: E402
+
 T = transforms.Compose(
     [
         v2.ToImage(),

tasks/class_medmnist/eval_class_dermamnist.py

@@ -1,34 +1,29 @@
 #!/usr/bin/env python3
-import sys
 import os
-
-root_dir = os.path.abspath(os.path.join(os.path.dirname(__file__), "../.."))
-sys.path.append(root_dir)
-
-from ncalab import ClassificationNCAModel, get_compute_device
+import sys
 
 import click
-
-from medmnist import INFO, DermaMNIST  # type: ignore[import-untyped]
-
 import torch  # type: ignore[import-untyped]
-from torchvision import transforms  # type: ignore[import-untyped]
-from torchvision.transforms import v2  # type: ignore[import-untyped]
-
 import torchmetrics
 import torchmetrics.classification
-
+from medmnist import INFO, DermaMNIST  # type: ignore[import-untyped]
+from torchvision import transforms  # type: ignore[import-untyped]
+from torchvision.transforms import v2  # type: ignore[import-untyped]
 from tqdm import tqdm
-
 from train_class_dermamnist import (
-    pad_noise,
-    alive_mask,
-    use_temporal_encoding,
-    fire_rate,
     WEIGHTS_PATH,
+    alive_mask,
     default_hidden_channels,
+    fire_rate,
+    pad_noise,
+    use_temporal_encoding,
 )
 
+root_dir = os.path.abspath(os.path.join(os.path.dirname(__file__), "../.."))
+sys.path.append(root_dir)
+
+from ncalab import ClassificationNCAModel, get_compute_device  # noqa: E402
+
 T = transforms.Compose(
     [
         v2.ToImage(),

tasks/class_medmnist/eval_class_pathmnist.py

@@ -1,34 +1,29 @@
 #!/usr/bin/env python3
-import sys
 import os
-
-root_dir = os.path.abspath(os.path.join(os.path.dirname(__file__), "../.."))
-sys.path.append(root_dir)
-
-from ncalab import ClassificationNCAModel, get_compute_device
+import sys
 
 import click
-
-from medmnist import INFO, PathMNIST  # type: ignore[import-untyped]
-
 import torch  # type: ignore[import-untyped]
-from torchvision import transforms  # type: ignore[import-untyped]
-from torchvision.transforms import v2  # type: ignore[import-untyped]
-
 import torchmetrics
 import torchmetrics.classification
-
+from medmnist import INFO, PathMNIST  # type: ignore[import-untyped]
+from torchvision import transforms  # type: ignore[import-untyped]
+from torchvision.transforms import v2  # type: ignore[import-untyped]
 from tqdm import tqdm
-
 from train_class_pathmnist import (
-    pad_noise,
-    alive_mask,
-    use_temporal_encoding,
-    fire_rate,
     WEIGHTS_PATH,
+    alive_mask,
     default_hidden_channels,
+    fire_rate,
+    pad_noise,
+    use_temporal_encoding,
 )
 
+root_dir = os.path.abspath(os.path.join(os.path.dirname(__file__), "../.."))
+sys.path.append(root_dir)
+
+from ncalab import ClassificationNCAModel, get_compute_device  # noqa: E402
+
 T = transforms.Compose(
     [
         v2.ToImage(),

tasks/class_medmnist/train_class_bloodmnist.py

@@ -1,26 +1,25 @@
 #!/usr/bin/env python3
-from pathlib import Path
-import sys
 import os
+import sys
+from pathlib import Path
+
+import click
+import torch  # type: ignore[import-untyped]
+from medmnist import INFO, BloodMNIST  # type: ignore[import-untyped]
+from torch.utils.tensorboard import SummaryWriter  # type: ignore[import-untyped]
+from torchvision import transforms  # type: ignore[import-untyped]
+from torchvision.transforms import v2  # type: ignore[import-untyped]
 
 root_dir = os.path.abspath(os.path.join(os.path.dirname(__file__), "../.."))
 sys.path.append(root_dir)
 
-from ncalab import (
-    ClassificationNCAModel,
+
+from ncalab import (  # noqa: E402
     BasicNCATrainer,
+    ClassificationNCAModel,
     get_compute_device,
 )
 
-import click
-
-from medmnist import INFO, BloodMNIST  # type: ignore[import-untyped]
-
-import torch  # type: ignore[import-untyped]
-from torchvision import transforms  # type: ignore[import-untyped]
-from torchvision.transforms import v2  # type: ignore[import-untyped]
-from torch.utils.tensorboard import SummaryWriter  # type: ignore[import-untyped]
-
 TASK_PATH = Path(__file__).parent.resolve()
 WEIGHTS_PATH = TASK_PATH / "weights"
 WEIGHTS_PATH.mkdir(exist_ok=True)

tasks/class_medmnist/train_class_dermamnist.py

@@ -1,28 +1,26 @@
 #!/usr/bin/env python3
-from pathlib import Path
-import sys
 import os
+import sys
+from pathlib import Path
+
+import click
+import numpy as np
+import torch  # type: ignore[import-untyped]
+from medmnist import INFO, DermaMNIST  # type: ignore[import-untyped]
+from torch.utils.tensorboard import SummaryWriter  # type: ignore[import-untyped]
+from torchvision import transforms  # type: ignore[import-untyped]
+from torchvision.transforms import v2  # type: ignore[import-untyped]
 
 root_dir = os.path.abspath(os.path.join(os.path.dirname(__file__), "../.."))
 sys.path.append(root_dir)
 
-from ncalab import (
-    ClassificationNCAModel,
+from ncalab import (  # noqa: E402
     BasicNCATrainer,
+    ClassificationNCAModel,
     get_compute_device,
     print_NCALab_banner,
 )
 
-import numpy as np
-import click
-
-from medmnist import INFO, DermaMNIST  # type: ignore[import-untyped]
-
-import torch  # type: ignore[import-untyped]
-from torchvision import transforms  # type: ignore[import-untyped]
-from torchvision.transforms import v2  # type: ignore[import-untyped]
-from torch.utils.tensorboard import SummaryWriter  # type: ignore[import-untyped]
-
 TASK_PATH = Path(__file__).parent.resolve()
 WEIGHTS_PATH = TASK_PATH / "weights"
 WEIGHTS_PATH.mkdir(exist_ok=True)

tasks/class_medmnist/train_class_pathmnist.py

@@ -1,26 +1,24 @@
 #!/usr/bin/env python3
-from pathlib import Path
-import sys
 import os
+import sys
+from pathlib import Path
+
+import click
+import torch  # type: ignore[import-untyped]
+from medmnist import INFO, PathMNIST  # type: ignore[import-untyped]
+from torch.utils.tensorboard import SummaryWriter  # type: ignore[import-untyped]
+from torchvision import transforms  # type: ignore[import-untyped]
+from torchvision.transforms import v2  # type: ignore[import-untyped]
 
 root_dir = os.path.abspath(os.path.join(os.path.dirname(__file__), "../.."))
 sys.path.append(root_dir)
 
-from ncalab import (
-    ClassificationNCAModel,
+from ncalab import (  # noqa: E402
     BasicNCATrainer,
+    ClassificationNCAModel,
     get_compute_device,
 )
 
-import click
-
-from medmnist import INFO, PathMNIST  # type: ignore[import-untyped]
-
-import torch  # type: ignore[import-untyped]
-from torchvision import transforms  # type: ignore[import-untyped]
-from torchvision.transforms import v2  # type: ignore[import-untyped]
-from torch.utils.tensorboard import SummaryWriter  # type: ignore[import-untyped]
-
 TASK_PATH = Path(__file__).parent.resolve()
 WEIGHTS_PATH = TASK_PATH / "weights"
 WEIGHTS_PATH.mkdir(exist_ok=True)

tasks/growing_emoji/README.md

@@ -2,3 +2,6 @@
 
 - Loosely based on the [Growing Cellular Automata](https://distill.pub/2020/growing-ca/) publication that sparked major interest in Neural Cellular Automata.
 - An evaluation script is provided to generate and show an image.
+
+
+## train_growing_emoji.py