brain-score.org submission (user:386) | (public:False)

brainscore_vision/models/mvimgnet_ss_04/__init__.py

@@ -0,0 +1,9 @@
+from brainscore_vision import model_registry
+from brainscore_vision.model_helpers.brain_transformation import ModelCommitment
+from .model import get_model, get_layers
+
+model_registry["mvimgnet_ss_04"] = lambda: ModelCommitment(
+    identifier="mvimgnet_ss_04",
+    activations_model=get_model("mvimgnet_ss_04"),
+    layers=get_layers("mvimgnet_ss_04"),
+)

brainscore_vision/models/mvimgnet_ss_04/model.py

@@ -0,0 +1,64 @@
+from brainscore_vision.model_helpers.check_submission import check_models
+import functools
+import os
+from urllib.request import urlretrieve
+import torchvision.models
+from brainscore_vision.model_helpers.activations.pytorch import PytorchWrapper
+from brainscore_vision.model_helpers.activations.pytorch import load_preprocess_images
+from pathlib import Path
+from brainscore_vision.model_helpers import download_weights
+import torch
+from collections import OrderedDict
+
+# This is an example implementation for submitting resnet-50 as a pytorch model
+
+# Attention: It is important, that the wrapper identifier is unique per model!
+# The results will otherwise be the same due to brain-scores internal result caching mechanism.
+# Please load your pytorch model for usage in CPU. There won't be GPUs available for scoring your model.
+# If the model requires a GPU, contact the brain-score team directly.
+from brainscore_vision.model_helpers.check_submission import check_models
+
+
+def get_model_list():
+    return ["mvimgnet_ss_04"]
+
+
+def get_model(name):
+    assert name == "mvimgnet_ss_04"
+    url = "https://users.flatironinstitute.org/~tyerxa/slow_steady/training_checkpoints/slow_steady/r2/LARS/lmda_0.4/latest-rank0.pt"
+    fh = urlretrieve(url)
+    state_dict = torch.load(fh[0], map_location=torch.device("cpu"))["state"]["model"]
+    model = load_composer_classifier(state_dict)
+    preprocessing = functools.partial(load_preprocess_images, image_size=224)
+    wrapper = PytorchWrapper(identifier=name, model=model, preprocessing=preprocessing)
+    wrapper.image_size = 224
+    return wrapper
+
+def load_composer_classifier(sd):
+    model = torchvision.models.resnet.resnet50()
+    new_sd = OrderedDict()
+    for k, v in sd.items():
+        if 'lin_cls' in k:
+            new_sd['fc.' + k.split('.')[-1]] = v
+        if ".f." not in k:
+            continue
+        parts = k.split(".")
+        idx = parts.index("f")
+        new_k = ".".join(parts[idx + 1 :])
+        new_sd[new_k] = v
+    model.load_state_dict(new_sd, strict=True)
+    return model
+
+def get_layers(name):
+    assert name == "mvimgnet_ss_04"
+
+    outs = ["layer1", "layer2", "layer3", "layer4"]
+    return outs
+
+
+def get_bibtex(model_identifier):
+    return """xx"""
+
+
+if __name__ == "__main__":
+    check_models.check_base_models(__name__)

brainscore_vision/models/mvimgnet_ss_04/setup.py

@@ -0,0 +1,25 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+
+from setuptools import setup, find_packages
+
+requirements = [ "torchvision",
+                 "torch"
+]
+
+setup(
+    packages=find_packages(exclude=['tests']),
+    include_package_data=True,
+    install_requires=requirements,
+    license="MIT license",
+    zip_safe=False,
+    keywords='brain-score template',
+    classifiers=[
+        'Development Status :: 2 - Pre-Alpha',
+        'Intended Audience :: Developers',
+        'License :: OSI Approved :: MIT License',
+        'Natural Language :: English',
+        'Programming Language :: Python :: 3.7',
+    ],
+    test_suite='tests',
+)

brainscore_vision/models/mvimgnet_ss_04/test.py

@@ -0,0 +1 @@
+# Left empty as part of 2023 models migration