add cifar_resnet18_1 to models

brainscore_vision/models/cifar_resnet18_1/__init__.py

@@ -0,0 +1,5 @@
+from brainscore_vision import model_registry
+from .model import get_model
+
+# Register the Barlow Twins model with custom weights
+model_registry['cifar_resnet18_1'] = lambda: get_model('cifar_resnet18_1')

brainscore_vision/models/cifar_resnet18_1/model.py

@@ -0,0 +1,68 @@
+import torch
+from pathlib import Path
+from torchvision.models import resnet18
+from brainscore_vision.model_helpers.activations.pytorch import PytorchWrapper
+from brainscore_vision.model_helpers.brain_transformation import ModelCommitment
+from brainscore_vision.model_helpers.activations.pytorch import load_preprocess_images
+from collections import OrderedDict
+from urllib.request import urlretrieve
+import functools
+import os
+
+
+# Custom model loader
+def get_model(name):
+    assert name == 'cifar_resnet18_1'
+    url = " https://www.dropbox.com/scl/fi/maqzcf3j87m7tp4sm1pab/barlow-cifar10-otu5cw89-ep-999.ckpt?rlkey=ou425fqbxxy6pe9lc4mz400mp&st=va93bqox&dl=1"
+    fh, _ = urlretrieve(url)
+    print(f"Downloaded weights file: {fh}, Size: {os.path.getsize(fh)} bytes")
+
+    checkpoint = torch.load(fh, map_location="cpu")
+    state_dict = checkpoint['state_dict']  # Adjust key if necessary
+    # Filter out projector layers
+    backbone_state_dict = {k.replace("backbone.", ""): v for k, v in state_dict.items() if not k.startswith("projector.")}
+    # Initialize ResNet18 backbone
+
+    model = resnet18(pretrained=False)
+    model.conv1 = torch.nn.Conv2d(3, 64, kernel_size=3, stride=1, padding=1, bias=False)
+    # print("First conv layer weights AFTER loading:")
+    # print(model.conv1.weight[0, 0, 0])
+    model.load_state_dict(backbone_state_dict, strict=False)
+    # print(f"Missing keys: {missing_keys}")
+    # print(f"Unexpected keys: {unexpected_keys}")
+    # print("First conv layer weights AFTER loading:")
+    # print(model.conv1.weight[0, 0, 0])
+    # print(model)
+
+
+    preprocessing = functools.partial(load_preprocess_images, image_size=224)
+
+    activations_model = PytorchWrapper(identifier='cifar_resnet18_1', model=model, preprocessing=preprocessing)
+
+
+    return ModelCommitment(
+        identifier='cifar_resnet18_1',
+        activations_model=activations_model,
+        layers=['layer1', 'layer2', 'layer3', 'layer4', 'avgpool']
+    )
+
+def get_model_list():
+    return ['cifar_resnet18_1']
+
+# Specify layers to test
+def get_layers(name):
+    assert name == 'cifar_resnet18_1'
+    return ['layer1', 'layer2', 'layer3', 'layer4', 'avgpool']
+
+def get_bibtex(model_identifier):
+    return """
+    @misc{resnet18_test_consistency,
+    title={ArtResNet18 Barlow Twins},
+    author={Claudia Noche},
+    year={2024},
+    }
+    """
+
+if __name__ == '__main__':
+    from brainscore_vision.model_helpers.check_submission import check_models
+    check_models.check_base_models(__name__)

brainscore_vision/models/cifar_resnet18_1/requirements.txt

@@ -0,0 +1,4 @@
+torch
+torchvision
+requests
+pathlib

brainscore_vision/models/cifar_resnet18_1/test.py

@@ -0,0 +1,10 @@
+import pytest
+import brainscore_vision
+
+@pytest.mark.travis_slow
+def test_cifar_resnet18_1():
+    model = brainscore_vision.load_model('cifar_resnet18_1')
+    assert model.identifier == 'cifar_resnet18_1'
+
+
+