brainscore-vision 2.2.1__py3-none-any.whl → 2.2.3__py3-none-any.whl

brainscore_vision/models/resnet50_z_axis_2/model.py

@@ -0,0 +1,200 @@
+
+from brainscore_vision.model_helpers.check_submission import check_models
+import functools
+import numpy as np
+import torch
+from brainscore_vision.model_helpers.activations.pytorch import PytorchWrapper
+from PIL import Image
+from torch import nn
+import pytorch_lightning as pl
+import torchvision.models as models
+import gdown
+import glob
+import os
+from brainscore_vision.model_helpers.activations.pytorch import load_preprocess_images
+
+def get_bibtex(model_identifier):
+    return 'VGG16'
+
+def get_model_list():
+    return ['resnet50_z_axis_iteration=2']
+
+def get_model(name):
+    keyword = 'z_axis'
+    iteration = 2
+    network = 'resnet50'
+    url = 'https://eggerbernhard.ch/shreya/latest_resnet50/z_axis_2.ckpt'
+    output = 'resnet50_z_axis_iteration=2.ckpt'
+    gdown.download(url, output)
+
+
+    if keyword != 'imagenet_trained' and keyword != 'no_training':
+        lx_whole = [f"resnet50_z_axis_iteration=2.ckpt"]
+        if len(lx_whole) > 1:
+            lx_whole = [lx_whole[-1]]
+    elif keyword == 'imagenet_trained' or keyword == 'no_training':
+        print('keyword is imagenet')
+        lx_whole = ['x']
+
+    for model_ckpt in lx_whole:
+        print(model_ckpt)
+        last_module_name = None
+        last_module = None
+        layers = []
+        if keyword == 'imagenet_trained' and network != 'clip':
+            model = torch.hub.load('pytorch/vision', network, pretrained=True)
+            for name, module in model.named_modules():
+                last_module_name = name
+                last_module = module
+                layers.append(name)
+        else:
+            model = torch.hub.load('pytorch/vision', network, pretrained=False)
+        if model_ckpt != 'x':
+            ckpt = torch.load(model_ckpt, map_location='cpu')
+        if model_ckpt != 'x' and network == 'alexnet' and keyword != 'imagenet_trained':
+            ckpt2 = {}
+            for keys in ckpt['state_dict']:
+                print(keys)
+                print(ckpt['state_dict'][keys].shape)
+                print('---')
+                k2 = keys.split('model.')[1]
+                ckpt2[k2] = ckpt['state_dict'][keys]
+            model.load_state_dict(ckpt2)
+        if model_ckpt != 'x' and network == 'vgg16' and keyword != 'imagenet_trained':
+            ckpt2 = {}
+            for keys in ckpt['state_dict']:
+                print(keys)
+                print(ckpt['state_dict'][keys].shape)
+                print('---')
+                k2 = keys.split('model.')[1]
+                ckpt2[k2] = ckpt['state_dict'][keys]
+            model.load_state_dict(ckpt2)
+        # Add more cases for other networks as needed
+    assert name == 'resnet50_z_axis_iteration=2'
+    url = 'https://eggerbernhard.ch/shreya/latest_resnet50/z_axis_2.ckpt'
+    output = 'resnet50_z_axis_iteration=2.ckpt'
+    gdown.download(url, output)
+    layers = []
+    for name, module in model._modules.items():
+        print(name, "->", module)
+        layers.append(name)
+
+    preprocessing = functools.partial(load_preprocess_images, image_size=224)
+    activations_model = PytorchWrapper(identifier=name, model=model, preprocessing=preprocessing)
+
+    return activations_model
+
+def get_layers(name):
+    keyword = 'z_axis'
+    iteration = 2
+    network = 'resnet50'
+    url = 'https://eggerbernhard.ch/shreya/latest_resnet50/z_axis_2.ckpt'
+    output = 'resnet50_z_axis_iteration=2.ckpt'
+    gdown.download(url, output)
+
+
+    if keyword != 'imagenet_trained' and keyword != 'no_training':
+        lx_whole = [f"resnet50_z_axis_iteration=2.ckpt"]
+        if len(lx_whole) > 1:
+            lx_whole = [lx_whole[-1]]
+    elif keyword == 'imagenet_trained' or keyword == 'no_training':
+        print('keyword is imagenet')
+        lx_whole = ['x']
+
+
+    for model_ckpt in lx_whole:
+        print(model_ckpt)
+        last_module_name = None
+        last_module = None
+        if keyword == 'imagenet_trained' and network != 'clip':
+            model = torch.hub.load('pytorch/vision', network, pretrained=True)
+            for name, module in model.named_modules():
+                last_module_name = name
+                last_module = module
+                layers.append(name)
+        else:
+            model = torch.hub.load('pytorch/vision', network, pretrained=False)
+        if model_ckpt != 'x':
+            ckpt = torch.load(model_ckpt, map_location='cpu')
+        if model_ckpt != 'x' and network == 'alexnet' and keyword != 'imagenet_trained':
+            ckpt2 = {}
+            for keys in ckpt['state_dict']:
+                print(keys)
+                print(ckpt['state_dict'][keys].shape)
+                print('---')
+                k2 = keys.split('model.')[1]
+                ckpt2[k2] = ckpt['state_dict'][keys]
+            model.load_state_dict(ckpt2)
+        if model_ckpt != 'x' and network == 'vgg16' and keyword != 'imagenet_trained':
+            ckpt2 = {}
+            for keys in ckpt['state_dict']:
+                print(keys)
+                print(ckpt['state_dict'][keys].shape)
+                print('---')
+                k2 = keys.split('model.')[1]
+                ckpt2[k2] = ckpt['state_dict'][keys]
+            model.load_state_dict(ckpt2)
+        # Add more cases for other networks as needed
+    layers = []
+    for name, module in model._modules.items():
+            print(name, "->", module)
+            layers.append(name)
+    return layers
+
+if __name__ == '__main__':
+    device = "cpu"
+    global model
+    global keyword
+    global network
+    global iteration
+    keyword = 'z_axis'
+    iteration = 2
+    network = 'resnet50'
+    url = 'https://eggerbernhard.ch/shreya/latest_resnet50/z_axis_2.ckpt'
+    output = 'resnet50_z_axis_iteration=2.ckpt'
+    gdown.download(url, output)
+
+
+    if keyword != 'imagenet_trained' and keyword != 'no_training':
+        lx_whole = [f"resnet50_z_axis_iteration=2.ckpt"]
+        if len(lx_whole) > 1:
+            lx_whole = [lx_whole[-1]]
+    elif keyword == 'imagenet_trained' or keyword == 'no_training':
+        print('keyword is imagenet')
+        lx_whole = ['x']
+
+    for model_ckpt in lx_whole:
+        print(model_ckpt)
+        last_module_name = None
+        last_module = None
+        layers = []
+        if keyword == 'imagenet_trained' and network != 'clip':
+            model = torch.hub.load('pytorch/vision', network, pretrained=True)
+            for name, module in model.named_modules():
+                last_module_name = name
+                last_module = module
+                layers.append(name)
+        else:
+            model = torch.hub.load('pytorch/vision', network, pretrained=False)
+        if model_ckpt != 'x':
+            ckpt = torch.load(model_ckpt, map_location='cpu')
+        if model_ckpt != 'x' and network == 'alexnet' and keyword != 'imagenet_trained':
+            ckpt2 = {}
+            for keys in ckpt['state_dict']:
+                print(keys)
+                print(ckpt['state_dict'][keys].shape)
+                print('---')
+                k2 = keys.split('model.')[1]
+                ckpt2[k2] = ckpt['state_dict'][keys]
+            model.load_state_dict(ckpt2)
+        if model_ckpt != 'x' and network == 'vgg16' and keyword != 'imagenet_trained':
+            ckpt2 = {}
+            for keys in ckpt['state_dict']:
+                print(keys)
+                print(ckpt['state_dict'][keys].shape)
+                print('---')
+                k2 = keys.split('model.')[1]
+                ckpt2[k2] = ckpt['state_dict'][keys]
+            model.load_state_dict(ckpt2)
+        # Add more cases for other networks as needed
+    check_models.check_base_models(__name__)

brainscore_vision/models/resnet50_z_axis_2/region_layer_map/resnet50_z_axis_iteration=2.json

@@ -0,0 +1,6 @@
+{
+    "V1": "maxpool",
+    "V2": "maxpool",
+    "V4": "maxpool",
+    "IT": "maxpool"
+}

brainscore_vision/models/resnet50_z_axis_2/setup.py

@@ -0,0 +1,29 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+
+from setuptools import setup, find_packages
+
+requirements = [
+    "torchvision",
+    "torch",
+    "gdown",
+    "pytorch_lightning",
+    "brainscore_vision"
+]
+
+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/resnet50_z_axis_2/test.py

@@ -0,0 +1,3 @@
+
+import pytest
+

brainscore_vision/models/resnet50_z_axis_3/__init__.py

@@ -0,0 +1,6 @@
+
+from brainscore_vision import model_registry
+from brainscore_vision.model_helpers.brain_transformation import ModelCommitment
+from .model import get_model, get_layers
+
+model_registry['resnet50_z_axis_iteration=3'] = lambda: ModelCommitment(identifier='resnet50_z_axis_iteration=3', activations_model=get_model('resnet50_z_axis_iteration=3'), layers=get_layers('resnet50_z_axis_iteration=3'))

brainscore_vision/models/resnet50_z_axis_3/model.py

@@ -0,0 +1,200 @@
+
+from brainscore_vision.model_helpers.check_submission import check_models
+import functools
+import numpy as np
+import torch
+from brainscore_vision.model_helpers.activations.pytorch import PytorchWrapper
+from PIL import Image
+from torch import nn
+import pytorch_lightning as pl
+import torchvision.models as models
+import gdown
+import glob
+import os
+from brainscore_vision.model_helpers.activations.pytorch import load_preprocess_images
+
+def get_bibtex(model_identifier):
+    return 'VGG16'
+
+def get_model_list():
+    return ['resnet50_z_axis_iteration=3']
+
+def get_model(name):
+    keyword = 'z_axis'
+    iteration = 3
+    network = 'resnet50'
+    url = 'https://eggerbernhard.ch/shreya/latest_resnet50/z_axis_3.ckpt'
+    output = 'resnet50_z_axis_iteration=3.ckpt'
+    gdown.download(url, output)
+
+
+    if keyword != 'imagenet_trained' and keyword != 'no_training':
+        lx_whole = [f"resnet50_z_axis_iteration=3.ckpt"]
+        if len(lx_whole) > 1:
+            lx_whole = [lx_whole[-1]]
+    elif keyword == 'imagenet_trained' or keyword == 'no_training':
+        print('keyword is imagenet')
+        lx_whole = ['x']
+
+    for model_ckpt in lx_whole:
+        print(model_ckpt)
+        last_module_name = None
+        last_module = None
+        layers = []
+        if keyword == 'imagenet_trained' and network != 'clip':
+            model = torch.hub.load('pytorch/vision', network, pretrained=True)
+            for name, module in model.named_modules():
+                last_module_name = name
+                last_module = module
+                layers.append(name)
+        else:
+            model = torch.hub.load('pytorch/vision', network, pretrained=False)
+        if model_ckpt != 'x':
+            ckpt = torch.load(model_ckpt, map_location='cpu')
+        if model_ckpt != 'x' and network == 'alexnet' and keyword != 'imagenet_trained':
+            ckpt2 = {}
+            for keys in ckpt['state_dict']:
+                print(keys)
+                print(ckpt['state_dict'][keys].shape)
+                print('---')
+                k2 = keys.split('model.')[1]
+                ckpt2[k2] = ckpt['state_dict'][keys]
+            model.load_state_dict(ckpt2)
+        if model_ckpt != 'x' and network == 'vgg16' and keyword != 'imagenet_trained':
+            ckpt2 = {}
+            for keys in ckpt['state_dict']:
+                print(keys)
+                print(ckpt['state_dict'][keys].shape)
+                print('---')
+                k2 = keys.split('model.')[1]
+                ckpt2[k2] = ckpt['state_dict'][keys]
+            model.load_state_dict(ckpt2)
+        # Add more cases for other networks as needed
+    assert name == 'resnet50_z_axis_iteration=3'
+    url = 'https://eggerbernhard.ch/shreya/latest_resnet50/z_axis_3.ckpt'
+    output = 'resnet50_z_axis_iteration=3.ckpt'
+    gdown.download(url, output)
+    layers = []
+    for name, module in model._modules.items():
+        print(name, "->", module)
+        layers.append(name)
+
+    preprocessing = functools.partial(load_preprocess_images, image_size=224)
+    activations_model = PytorchWrapper(identifier=name, model=model, preprocessing=preprocessing)
+
+    return activations_model
+
+def get_layers(name):
+    keyword = 'z_axis'
+    iteration = 3
+    network = 'resnet50'
+    url = 'https://eggerbernhard.ch/shreya/latest_resnet50/z_axis_3.ckpt'
+    output = 'resnet50_z_axis_iteration=3.ckpt'
+    gdown.download(url, output)
+
+
+    if keyword != 'imagenet_trained' and keyword != 'no_training':
+        lx_whole = [f"resnet50_z_axis_iteration=3.ckpt"]
+        if len(lx_whole) > 1:
+            lx_whole = [lx_whole[-1]]
+    elif keyword == 'imagenet_trained' or keyword == 'no_training':
+        print('keyword is imagenet')
+        lx_whole = ['x']
+
+
+    for model_ckpt in lx_whole:
+        print(model_ckpt)
+        last_module_name = None
+        last_module = None
+        if keyword == 'imagenet_trained' and network != 'clip':
+            model = torch.hub.load('pytorch/vision', network, pretrained=True)
+            for name, module in model.named_modules():
+                last_module_name = name
+                last_module = module
+                layers.append(name)
+        else:
+            model = torch.hub.load('pytorch/vision', network, pretrained=False)
+        if model_ckpt != 'x':
+            ckpt = torch.load(model_ckpt, map_location='cpu')
+        if model_ckpt != 'x' and network == 'alexnet' and keyword != 'imagenet_trained':
+            ckpt2 = {}
+            for keys in ckpt['state_dict']:
+                print(keys)
+                print(ckpt['state_dict'][keys].shape)
+                print('---')
+                k2 = keys.split('model.')[1]
+                ckpt2[k2] = ckpt['state_dict'][keys]
+            model.load_state_dict(ckpt2)
+        if model_ckpt != 'x' and network == 'vgg16' and keyword != 'imagenet_trained':
+            ckpt2 = {}
+            for keys in ckpt['state_dict']:
+                print(keys)
+                print(ckpt['state_dict'][keys].shape)
+                print('---')
+                k2 = keys.split('model.')[1]
+                ckpt2[k2] = ckpt['state_dict'][keys]
+            model.load_state_dict(ckpt2)
+        # Add more cases for other networks as needed
+    layers = []
+    for name, module in model._modules.items():
+            print(name, "->", module)
+            layers.append(name)
+    return layers
+
+if __name__ == '__main__':
+    device = "cpu"
+    global model
+    global keyword
+    global network
+    global iteration
+    keyword = 'z_axis'
+    iteration = 3
+    network = 'resnet50'
+    url = 'https://eggerbernhard.ch/shreya/latest_resnet50/z_axis_3.ckpt'
+    output = 'resnet50_z_axis_iteration=3.ckpt'
+    gdown.download(url, output)
+
+
+    if keyword != 'imagenet_trained' and keyword != 'no_training':
+        lx_whole = [f"resnet50_z_axis_iteration=3.ckpt"]
+        if len(lx_whole) > 1:
+            lx_whole = [lx_whole[-1]]
+    elif keyword == 'imagenet_trained' or keyword == 'no_training':
+        print('keyword is imagenet')
+        lx_whole = ['x']
+
+    for model_ckpt in lx_whole:
+        print(model_ckpt)
+        last_module_name = None
+        last_module = None
+        layers = []
+        if keyword == 'imagenet_trained' and network != 'clip':
+            model = torch.hub.load('pytorch/vision', network, pretrained=True)
+            for name, module in model.named_modules():
+                last_module_name = name
+                last_module = module
+                layers.append(name)
+        else:
+            model = torch.hub.load('pytorch/vision', network, pretrained=False)
+        if model_ckpt != 'x':
+            ckpt = torch.load(model_ckpt, map_location='cpu')
+        if model_ckpt != 'x' and network == 'alexnet' and keyword != 'imagenet_trained':
+            ckpt2 = {}
+            for keys in ckpt['state_dict']:
+                print(keys)
+                print(ckpt['state_dict'][keys].shape)
+                print('---')
+                k2 = keys.split('model.')[1]
+                ckpt2[k2] = ckpt['state_dict'][keys]
+            model.load_state_dict(ckpt2)
+        if model_ckpt != 'x' and network == 'vgg16' and keyword != 'imagenet_trained':
+            ckpt2 = {}
+            for keys in ckpt['state_dict']:
+                print(keys)
+                print(ckpt['state_dict'][keys].shape)
+                print('---')
+                k2 = keys.split('model.')[1]
+                ckpt2[k2] = ckpt['state_dict'][keys]
+            model.load_state_dict(ckpt2)
+        # Add more cases for other networks as needed
+    check_models.check_base_models(__name__)

brainscore_vision/models/resnet50_z_axis_3/region_layer_map/resnet50_z_axis_iteration=3.json

@@ -0,0 +1,6 @@
+{
+    "V1": "maxpool",
+    "V2": "layer2",
+    "V4": "maxpool",
+    "IT": "layer1"
+}

brainscore_vision/models/resnet50_z_axis_3/setup.py

@@ -0,0 +1,29 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+
+from setuptools import setup, find_packages
+
+requirements = [
+    "torchvision",
+    "torch",
+    "gdown",
+    "pytorch_lightning",
+    "brainscore_vision"
+]
+
+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/resnet50_z_axis_3/test.py

@@ -0,0 +1,3 @@
+
+import pytest
+

brainscore_vision/models/resnet50_z_axis_5/__init__.py

@@ -0,0 +1,6 @@
+
+from brainscore_vision import model_registry
+from brainscore_vision.model_helpers.brain_transformation import ModelCommitment
+from .model import get_model, get_layers
+
+model_registry['resnet50_z_axis_iteration=5'] = lambda: ModelCommitment(identifier='resnet50_z_axis_iteration=5', activations_model=get_model('resnet50_z_axis_iteration=5'), layers=get_layers('resnet50_z_axis_iteration=5'))