Merge pull request #73 from dwoiwode/image_classification_full_cs

Image classification full cs

Author: LMZimmer

autoPyTorch/components/networks/activations.py

@@ -19,5 +19,5 @@ def get_activation(name, inplace=False):
     if name not in all_activations:
         raise ValueError('Activation ' + str(name) + ' not defined')
     activation = all_activations[name]
-    activation_kwargs = { 'inplace': True } if 'inplace' in inspect.getargspec(activation)[0] else dict()
+    activation_kwargs = { 'inplace': True } if 'inplace' in inspect.getfullargspec(activation)[0] else dict()
     return activation(**activation_kwargs)

autoPyTorch/pipeline/nodes/image/create_image_dataloader.py

@@ -18,6 +18,9 @@
 from torch.utils.data.sampler import SubsetRandomSampler
 from torchvision import datasets, models, transforms
 
+from autoPyTorch.utils.transforms import transform_int64
+
+
 class CreateImageDataLoader(CreateDataLoader):
 
     def fit(self, pipeline_config, hyperparameter_config, X, Y, train_indices, valid_indices, train_transform, valid_transform, dataset_info):
@@ -27,18 +30,19 @@ def fit(self, pipeline_config, hyperparameter_config, X, Y, train_indices, valid
 
         torch.manual_seed(pipeline_config["random_seed"])
         hyperparameter_config = ConfigWrapper(self.get_name(), hyperparameter_config)
+        to_int64 = transform_int64
 
         if dataset_info.default_dataset:
             train_dataset = dataset_info.default_dataset(root=pipeline_config['default_dataset_download_dir'], train=True, download=True, transform=train_transform)
             if valid_indices is not None:
                 valid_dataset = dataset_info.default_dataset(root=pipeline_config['default_dataset_download_dir'], train=True, download=True, transform=valid_transform)
         elif len(X.shape) > 1:
-            train_dataset = XYDataset(X, Y, transform=train_transform, target_transform=lambda y: y.astype(np.int64))
-            valid_dataset = XYDataset(X, Y, transform=valid_transform, target_transform=lambda y: y.astype(np.int64))
+            train_dataset = XYDataset(X, Y, transform=train_transform, target_transform=to_int64)
+            valid_dataset = XYDataset(X, Y, transform=valid_transform, target_transform=to_int64)
         else:
-            train_dataset = ImageFilelist(X, Y, transform=train_transform, target_transform=lambda y: y.astype(np.int64), cache_size=pipeline_config['dataloader_cache_size_mb'] * 1000, image_size=dataset_info.x_shape[2:])
+            train_dataset = ImageFilelist(X, Y, transform=train_transform, target_transform=to_int64, cache_size=pipeline_config['dataloader_cache_size_mb'] * 1000, image_size=dataset_info.x_shape[2:])
             if valid_indices is not None:
-                valid_dataset = ImageFilelist(X, Y, transform=valid_transform, target_transform=lambda y: y.astype(np.int64), cache_size=0, image_size=dataset_info.x_shape[2:])
+                valid_dataset = ImageFilelist(X, Y, transform=valid_transform, target_transform=to_int64, cache_size=0, image_size=dataset_info.x_shape[2:])
                 valid_dataset.cache = train_dataset.cache
 
         train_loader = DataLoader(

autoPyTorch/utils/transforms.py

@@ -0,0 +1,5 @@
+import numpy as np
+
+
+def transform_int64(y):
+    return y.astype(np.int64)

examples/basics/image_classification.py

@@ -0,0 +1,33 @@
+__license__ = "BSD"
+
+import os, sys
+import numpy as np
+
+sys.path.append(os.path.abspath(os.path.join(__file__, "..", "..", "..")))
+from autoPyTorch import AutoNetImageClassification
+
+# Note: You can write your own datamanager! Call fit with respective train, valid data (numpy matrices) 
+csv_dir = os.path.abspath("../../datasets/example.csv")
+
+def main():
+    X_train = np.array([csv_dir])
+    Y_train = np.array([0])
+
+    # Note: every parameter has a default value, you do not have to specify anything. The given parameter allow a fast test.
+    autonet = AutoNetImageClassification(config_preset="full_cs", result_logger_dir="logs/")
+
+    res = autonet.fit(X_train=X_train,
+                      Y_train=Y_train,
+                      images_shape=[3, 32, 32],
+                      min_budget=600,
+                      max_budget=900,
+                      max_runtime=1800,
+                      save_checkpoints=True,
+                      images_root_folders=[os.path.abspath("../../datasets/example_images")])
+
+    print(res)
+    print("Score:", autonet.score(X_test=X_train, Y_test=Y_train))
+
+
+if __name__ == '__main__':
+    main()