Merge pull request #4 from openml/feature/dataloader_imagedataset

Feature/dataloader image dataset

Author: Taniya-Das

.gitignore

@@ -4,3 +4,7 @@ openml_pytorch/__pycache__/
 .idea/
 
 openml_pytorch/layers/__pycache__/
+
+venv
+
+model.onnx

README.md

@@ -32,4 +32,10 @@ run = openml.runs.run_model_on_task(model, task, avoid_duplicate_runs=False)
 run.publish()
 print('URL for run: %s/run/%d' % (openml.config.server, run.run_id))
 ```
-Note: The input layer of the network should be compatible with OpenML data output shape. Please check examples for more information.
+Note: The input layer of the network should be compatible with OpenML data output shape. Please check [examples](/examples/) for more information.
+
+Additionally, if you want to publish the run with onnx file, then you must call ```openml_pytorch.add_onnx_to_run()``` immediately before ```run.publish()```. 
+
+```python
+run = openml_pytorch.add_onnx_to_run(run)
+```

examples/create_new_task.py

@@ -0,0 +1,38 @@
+import openml
+from openml.tasks import OpenMLClassificationTask
+
+task = openml.tasks.get_task(361175)
+# openml.config.apikey = 'KEY'
+# Define task parameters
+task_type = openml.tasks.TaskType.SUPERVISED_CLASSIFICATION
+evaluation_measure = 'predictive_accuracy'
+estimation_procedure = {
+    'type': 'crossvalidation',
+    'parameters': {
+        'number_repeats': '1',
+        'number_folds': '10',
+        'percentage': '',
+        'stratified_sampling': 'true'
+    },
+    'data_splits_url': 'https://api.openml.org/api_splits/get/361175/Task_361175_splits.arff'
+}
+target_name = 'CATEGORY'
+class_labels = ['Adrenal_gland', 'Bile-duct', 'Bladder', 'Breast', 'Cervix', 'Colon', 'Esophagus', 'HeadNeck', 'Kidney', 'Liver', 'Lung', 'Ovarian', 'Pancreatic', 'Prostate', 'Skin', 'Stomach', 'Testis', 'Thyroid', 'Uterus']
+cost_matrix = None
+
+# 'split': <openml.tasks.split.OpenMLSplit object at 0x7efca59476a0>
+
+# Create the task
+new_task = openml.tasks.create_task(
+    task_type=task_type,
+    dataset_id=task.dataset_id, 
+    estimation_procedure_id = task.estimation_procedure_id,
+    # estimation_procedure=estimation_procedure,
+    target_name=target_name,
+    class_labels=class_labels,
+    cost_matrix=cost_matrix
+)
+
+print(new_task)
+
+new_task.publish()

examples/pytorch_IndoorScenes_classification.py

@@ -0,0 +1,65 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+import openml_pytorch
+import openml
+import warnings
+
+# warnings.simplefilter(action='ignore', category=FutureWarning)
+warnings.simplefilter(action='ignore')
+
+
+def evaluate_torch_model(model):
+    # Download CV splits
+    task = openml.tasks.get_task(362070)
+    # Evaluate model
+    run = openml.runs.run_model_on_task(model, task, avoid_duplicate_runs=False)
+    # Publish
+    run = openml_pytorch.add_onnx_to_run(run) # Optional, to inspect afterward
+    run.publish()
+    return run
+
+from torchvision import models
+from torchvision.transforms import v2
+
+class Model2(nn.Module):
+    def __init__(self, num_classes=67):
+        super(Model2, self).__init__()
+        self.conv1 = nn.Conv2d(3, 6, 5)
+        self.pool = nn.MaxPool2d(2, 2)
+        self.conv2 = nn.Conv2d(6, 16, 5)
+        self.fc1 = nn.Linear(13456, 120)
+        self.fc2 = nn.Linear(120, 84)
+        self.fc3 = nn.Linear(84, num_classes) 
+
+    def forward(self, x):
+        x = self.pool(F.relu(self.conv1(x)))
+        x = self.pool(F.relu(self.conv2(x)))
+        x = torch.flatten(x, 1) # flatten all dimensions except batch
+        x = F.relu(self.fc1(x))
+        x = F.relu(self.fc2(x))
+        x = self.fc3(x)
+        return x
+
+# Training parameters
+openml_pytorch.config.batch_size = 32
+openml_pytorch.config.epoch_count = 1
+openml_pytorch.config.image_size = 128
+
+transforms = v2.Compose([
+    v2.RandomResizedCrop(size=(224, 224), antialias=True),
+    v2.RandomHorizontalFlip(p=0.5),
+    v2.ToDtype(torch.float32, scale=True),
+    v2.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
+])
+
+openml_pytorch.config.data_augemntation = transforms 
+openml_pytorch.config.perform_validation = True
+
+openml.config.apikey = 'key'
+openml_pytorch.config.file_dir = openml.config.get_cache_directory()+'/datasets/45923/Images/'
+openml_pytorch.config.filename_col = "Filename"
+
+# Run
+run = evaluate_torch_model(Model2()) # Replace with your model
+print('URL for run: %s/run/%d?api_key=%s' % (openml.config.server, run.run_id, openml.config.apikey)) 

examples/pytorch_image_classification_example.py

@@ -0,0 +1,98 @@
+"""
+PyTorch image classification model example
+==================
+
+An example of a pytorch network that classifies meta album images.
+"""
+
+import torch.nn
+import torch.optim
+
+import openml
+import openml_pytorch
+import openml_pytorch.layers
+import openml_pytorch.config
+import logging
+
+import warnings
+import pandas as pd
+
+# Suppress FutureWarning messages
+warnings.simplefilter(action='ignore')
+
+############################################################################
+# Enable logging in order to observe the progress while running the example.
+openml.config.logger.setLevel(logging.DEBUG)
+openml_pytorch.config.logger.setLevel(logging.DEBUG)
+############################################################################
+
+############################################################################
+import torch.nn as nn
+import torch.nn.functional as F
+
+# Example model. You can do better :)
+class Net(nn.Module):
+    def __init__(self):
+        super().__init__()
+        self.conv1 = nn.Conv2d(3, 6, 5)
+        self.pool = nn.MaxPool2d(2, 2)
+        self.conv2 = nn.Conv2d(6, 16, 5)
+        self.fc1 = nn.Linear(13456, 120)
+        self.fc2 = nn.Linear(120, 84)
+        self.fc3 = nn.Linear(84, 19) # To user - Remember to set correct size of last layer. 
+
+    def forward(self, x):
+        x = self.pool(F.relu(self.conv1(x)))
+        x = self.pool(F.relu(self.conv2(x)))
+        x = torch.flatten(x, 1) # flatten all dimensions except batch
+        x = F.relu(self.fc1(x))
+        x = F.relu(self.fc2(x))
+        x = self.fc3(x)
+        return x
+
+net = Net()
+
+############################################################################
+openml.config.apikey = 'key'
+openml_pytorch.config.file_dir = openml.config.get_cache_directory()+'/datasets/44312/PNU_Micro/images/'
+openml_pytorch.config.filename_col = "FILE_NAME"
+openml_pytorch.config.perform_validation = False
+############################################################################
+# The main network, composed of the above specified networks.
+model = net
+
+############################################################################
+# Download the OpenML task for the Meta_Album_PNU_Micro dataset.
+task = openml.tasks.get_task(361987)
+
+############################################################################
+# Run the model on the task (requires an API key).m
+run = openml.runs.run_model_on_task(model, task, avoid_duplicate_runs=False)
+
+# If you want to publish the run with the onnx file, 
+# then you must call openml_pytorch.add_onnx_to_run() immediately before run.publish(). 
+# When you publish, onnx file of last trained model is uploaded. 
+# Careful to not call this function when another run_model_on_task is called in between, 
+# as during publish later, only the last trained model (from last run_model_on_task call) is uploaded.   
+run = openml_pytorch.add_onnx_to_run(run)
+
+run.publish()
+
+print('URL for run: %s/run/%d' % (openml.config.server, run.run_id))
+############################################################################
+
+# Visualize model in netron
+
+from urllib.request import urlretrieve
+
+published_run = openml.runs.get_run(run.run_id)
+url = 'https://api.openml.org/data/download/{}/model.onnx'.format(published_run.output_files['onnx_model'])
+
+file_path, _ = urlretrieve(url, 'model.onnx')
+
+import netron
+# Visualize the ONNX model using Netron
+netron.start(file_path)
+
+
+

examples/pytorch_pretrained_image_classification_example.py

@@ -0,0 +1,87 @@
+"""
+PyTorch image classification model using pre-trained ResNet model example
+==================
+
+An example of a pytorch network that classifies meta album images.
+"""
+
+import torch.nn
+import torch.optim
+
+import openml
+import openml_pytorch
+import openml_pytorch.layers
+import openml_pytorch.config
+import logging
+
+############################################################################
+# Enable logging in order to observe the progress while running the example.
+openml.config.logger.setLevel(logging.DEBUG)
+openml_pytorch.config.logger.setLevel(logging.DEBUG)
+############################################################################
+
+############################################################################
+import torch.nn as nn
+import torch.nn.functional as F
+
+# Example model. You can do better :)
+import torchvision.models as models
+
+# Load the pre-trained ResNet model
+model = models.resnet50(pretrained=True)
+
+# Modify the last fully connected layer to the required number of classes
+num_classes = 20
+in_features = model.fc.in_features
+model.fc = nn.Linear(in_features, num_classes)
+
+# Optional: If you're fine-tuning, you may want to freeze the pre-trained layers
+for param in model.parameters():
+    param.requires_grad = False
+
+# If you want to train the last layer only (the newly added layer)
+for param in model.fc.parameters():
+    param.requires_grad = True
+
+############################################################################
+# Setting an appropriate optimizer 
+from openml import OpenMLTask
+
+def custom_optimizer_gen(model: torch.nn.Module, task: OpenMLTask) -> torch.optim.Optimizer:
+    return torch.optim.Adam(model.fc.parameters())
+
+openml_pytorch.config.optimizer_gen = custom_optimizer_gen
+
+############################################################################
+
+# openml.config.apikey = 'KEY'
+
+############################################################################
+# Download the OpenML task for the Meta_Album_PNU_Micro dataset.
+task = openml.tasks.get_task(361152)
+
+############################################################################
+# Run the model on the task (requires an API key).m
+run = openml.runs.run_model_on_task(model, task, avoid_duplicate_runs=False)
+
+# Publish the experiment on OpenML (optional, requires an API key).
+run.publish()
+
+print('URL for run: %s/run/%d' % (openml.config.server, run.run_id))
+
+############################################################################
+
+# Visualize model in netron
+import netron
+
+# Define input size
+input_size = (32,3,128,128)
+
+# Create a dummy input with the specified size
+dummy_input = torch.randn(input_size)
+
+# Export the model to ONNX
+torch.onnx.export(model, dummy_input, "model.onnx", verbose=True)
+
+# Visualize the ONNX model using Netron
+netron.start("model.onnx")