# Torch Model Manager
**Torch Model Manager** is an open-source python project designed for Deep Learning developpers that aims to make the use of pytorch library easy. The version  is still under developpment. The package allows us to access, search and delete layers by index, attributes or instance.
### Examples of Use
1. **Initialization**
```python
from torchvision import
from torch_model_manager import TorchModelManager
# Assume you have a PyTorch model 'model'
model = models.vgg16(pretrained=True)
model_manager = TorchModelManager(model)
```
2. **Get Named Layers**
```python
named_layers = model_manager.get_named_layers()
# output
>>> {
'features': {
'0': 'Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))',
'1': 'ReLU(inplace=True)',
'2': 'Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))',
'3': 'ReLU(inplace=True)',
'4': 'MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)',
'5': 'Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))',
'6': 'ReLU(inplace=True)',
'7': 'Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))',
'8': 'ReLU(inplace=True)',
'9': 'MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)',
'10': 'Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))',
'11': 'ReLU(inplace=True)',
'12': 'Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))',
'13': 'ReLU(inplace=True)',
'14': 'Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))',
'15': 'ReLU(inplace=True)',
'16': 'MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)',
'17': 'Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))',
'18': 'ReLU(inplace=True)',
'19': 'Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))',
'20': 'ReLU(inplace=True)',
'21': 'Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))',
'22': 'ReLU(inplace=True)',
'23': 'MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)',
'24': 'Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))',
'25': 'ReLU(inplace=True)',
'26': 'Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))',
'27': 'ReLU(inplace=True)',
'28': 'Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))',
'29': 'ReLU(inplace=True)',
'30': 'MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)',
},
'avgpool': 'AdaptiveAvgPool2d(output_size=(7, 7))',
'classifier': {
'0': 'Linear(in_features=25088, out_features=4096, bias=True)',
'1': 'ReLU(inplace=True)',
'2': 'Dropout(p=0.5, inplace=False)',
'3': 'Linear(in_features=4096, out_features=4096, bias=True)',
'4': 'ReLU(inplace=True)',
'5': 'Dropout(p=0.5, inplace=False)',
'6': 'Linear(in_features=4096, out_features=1000, bias=True)'
}
}
```
This method allows the user to access the overall architecture of the model in dictionnary format.
3. **Get Layer by Index**
```python
layer_index = ['classifier', 6]
layer = model_manager.get_layer_by_index(layer_index)
>>> Linear(in_features=4096, out_features=1000, bias=True)
```
The index is represented by a list, where each position represents a level. For instance, in the previous example, 'classifier' is the index to access the first level of the model architecture, and 6 is the index of the layer at the second level.
4. **Get Layer by Attribute**
```python
layers = model_manager.get_layer_by_attribute('out_features', 1000, '==')
>>> {('classifier', 6): Linear(in_features=4096, out_features=1000, bias=True)}
```
Retrieves all the layers satifying the given condition `out_features = 1000`.
5. **Get Layers by Conditions**
```python
# Retrieve layers that satisfy the given conditions
conditions = {
'and': [{'==': ('kernel_size', (1, 1))}, {'==': ('stride', (1, 1))}],
'or': [{'==': ('kernel_size', (3, 3))}]
}
layers = model_manager.get_layer_by_attributes(conditions)
>>> {('features', 0): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 5): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 7): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 10): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 12): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 14): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 17): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 19): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 21): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 24): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 26): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 28): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))}
```
It is almost the same as the previous one, but this time it extracts the layers that satisfy a set of conditions.
6. **Get Layer by Instance**
```python
# Search for layers in the model by their instance type
layers = model_manager.get_layer_by_instance(nn.Conv2d)
>>> {('features', 0): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 5): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 7): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 10): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 12): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 14): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 17): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 19): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 21): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 24): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 26): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 28): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))}
```
The `get_layer_by_instance` method allows you to extract layers of a specific type from the model. In the previous example, the extracted layers are convolutional layers.
7. **Delete Layer by Index**
The deletion process involves the following steps:
1. Search for the layers and retrieve their indexes.
2. Delete the layers at the corresponding indexes.
Here is an example of how to delete layers using different methods:
- Delete a layer by index:
```python
# Delete a layer from the model using its index
model_manager.delete_layer_by_index(['features', 0])
```
8. **Delete Layer by Attribute**
```python
# Delete layers from the model based on a specific attribute
model_manager.delete_layer_by_attribute('activation', 'relu', '==')
```
9. **Delete Layers by Conditions**
```python
# Delete layers from the model based on multiple conditions
conditions = {
'and': [{'==': ('kernel_size', (1, 1))}, {'==': ('stride', (1, 1))}],
'or': [{'==': ('kernel_size', (3, 3))}]
}
model_manager.delete_layer_by_attributes(conditions)
```
10. **Delete Layer by Instance**
```python
# Delete layers from the model by their instance type
model_manager.delete_layer_by_instance(nn.Conv2d)
```
Raw data
{
"_id": null,
"home_page": "https://github.com/Billal-MOKHTARI/torch-model-manager",
"name": "torch-model-manager",
"maintainer": null,
"docs_url": null,
"requires_python": null,
"maintainer_email": null,
"keywords": "PyTorch, Deep Learning, Machine Learning, High Level Programming",
"author": "Billal MOKHTARI",
"author_email": "mokhtaribillal1@gmail.com",
"download_url": "https://files.pythonhosted.org/packages/d8/36/8336c07cf05cc134b5135cb7218b6d083d68321e82b592ca5d53497f236f/torch_model_manager-0.2.0.tar.gz",
"platform": null,
"description": "# Torch Model Manager\n\n\n**Torch Model Manager** is an open-source python project designed for Deep Learning developpers that aims to make the use of pytorch library easy. The version  is still under developpment. The package allows us to access, search and delete layers by index, attributes or instance.\n\n### Examples of Use\n1. **Initialization**\n```python\nfrom torchvision import\nfrom torch_model_manager import TorchModelManager\n\n# Assume you have a PyTorch model 'model'\nmodel = models.vgg16(pretrained=True)\n\nmodel_manager = TorchModelManager(model)\n```\n\n2. **Get Named Layers**\n```python\nnamed_layers = model_manager.get_named_layers()\n\n# output\n>>> {\n 'features': {\n '0': 'Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))',\n '1': 'ReLU(inplace=True)',\n '2': 'Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))',\n '3': 'ReLU(inplace=True)',\n '4': 'MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)',\n '5': 'Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))',\n '6': 'ReLU(inplace=True)',\n '7': 'Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))',\n '8': 'ReLU(inplace=True)',\n '9': 'MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)',\n '10': 'Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))',\n '11': 'ReLU(inplace=True)',\n '12': 'Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))',\n '13': 'ReLU(inplace=True)',\n '14': 'Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))',\n '15': 'ReLU(inplace=True)',\n '16': 'MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)',\n '17': 'Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))',\n '18': 'ReLU(inplace=True)',\n '19': 'Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))',\n '20': 'ReLU(inplace=True)',\n '21': 'Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))',\n '22': 'ReLU(inplace=True)',\n '23': 'MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)',\n '24': 'Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))',\n '25': 'ReLU(inplace=True)',\n '26': 'Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))',\n '27': 'ReLU(inplace=True)',\n '28': 'Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))',\n '29': 'ReLU(inplace=True)',\n '30': 'MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)',\n },\n 'avgpool': 'AdaptiveAvgPool2d(output_size=(7, 7))',\n 'classifier': {\n '0': 'Linear(in_features=25088, out_features=4096, bias=True)',\n '1': 'ReLU(inplace=True)',\n '2': 'Dropout(p=0.5, inplace=False)',\n '3': 'Linear(in_features=4096, out_features=4096, bias=True)',\n '4': 'ReLU(inplace=True)',\n '5': 'Dropout(p=0.5, inplace=False)',\n '6': 'Linear(in_features=4096, out_features=1000, bias=True)'\n }\n}\n\n```\n\nThis method allows the user to access the overall architecture of the model in dictionnary format.\n\n\n3. **Get Layer by Index**\n```python\nlayer_index = ['classifier', 6]\nlayer = model_manager.get_layer_by_index(layer_index)\n\n>>> Linear(in_features=4096, out_features=1000, bias=True)\n```\n\nThe index is represented by a list, where each position represents a level. For instance, in the previous example, 'classifier' is the index to access the first level of the model architecture, and 6 is the index of the layer at the second level.\n\n4. **Get Layer by Attribute**\n```python\nlayers = model_manager.get_layer_by_attribute('out_features', 1000, '==')\n>>> {('classifier', 6): Linear(in_features=4096, out_features=1000, bias=True)}\n```\nRetrieves all the layers satifying the given condition `out_features = 1000`.\n\n5. **Get Layers by Conditions**\n```python\n# Retrieve layers that satisfy the given conditions\nconditions = {\n 'and': [{'==': ('kernel_size', (1, 1))}, {'==': ('stride', (1, 1))}],\n 'or': [{'==': ('kernel_size', (3, 3))}]\n }\nlayers = model_manager.get_layer_by_attributes(conditions)\n>>> {('features', 0): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 5): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 7): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 10): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 12): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 14): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 17): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 19): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 21): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 24): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 26): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 28): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))}\n```\n\nIt is almost the same as the previous one, but this time it extracts the layers that satisfy a set of conditions.\n\n6. **Get Layer by Instance**\n```python\n# Search for layers in the model by their instance type\nlayers = model_manager.get_layer_by_instance(nn.Conv2d)\n>>> {('features', 0): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 5): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 7): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 10): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 12): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 14): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 17): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 19): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 21): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 24): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 26): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)), ('features', 28): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))}\n```\n\nThe `get_layer_by_instance` method allows you to extract layers of a specific type from the model. In the previous example, the extracted layers are convolutional layers.\n\n7. **Delete Layer by Index**\nThe deletion process involves the following steps:\n\n1. Search for the layers and retrieve their indexes.\n2. Delete the layers at the corresponding indexes.\n\nHere is an example of how to delete layers using different methods:\n\n- Delete a layer by index:\n\n```python\n# Delete a layer from the model using its index\nmodel_manager.delete_layer_by_index(['features', 0])\n```\n\n8. **Delete Layer by Attribute**\n```python\n# Delete layers from the model based on a specific attribute\nmodel_manager.delete_layer_by_attribute('activation', 'relu', '==')\n```\n9. **Delete Layers by Conditions**\n```python\n# Delete layers from the model based on multiple conditions\nconditions = {\n 'and': [{'==': ('kernel_size', (1, 1))}, {'==': ('stride', (1, 1))}],\n 'or': [{'==': ('kernel_size', (3, 3))}]\n}\nmodel_manager.delete_layer_by_attributes(conditions)\n```\n10. **Delete Layer by Instance**\n\n```python\n# Delete layers from the model by their instance type\nmodel_manager.delete_layer_by_instance(nn.Conv2d)\n```\n\n",
"bugtrack_url": null,
"license": null,
"summary": "A package for managing PyTorch models",
"version": "0.2.0",
"project_urls": {
"Homepage": "https://github.com/Billal-MOKHTARI/torch-model-manager"
},
"split_keywords": [
"pytorch",
" deep learning",
" machine learning",
" high level programming"
],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "019d3ab344bcc6abe18e5d2964e2daa8edfaf1362814192e1b9fed015b11e08c",
"md5": "70a1c0ab876a6a5d65cbf1f8eb9e8e27",
"sha256": "f7e728de90c25364191753a06348170f7f7207f6a5523422bd5e8cb653c84f51"
},
"downloads": -1,
"filename": "torch_model_manager-0.2.0-py3-none-any.whl",
"has_sig": false,
"md5_digest": "70a1c0ab876a6a5d65cbf1f8eb9e8e27",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": null,
"size": 14628,
"upload_time": "2024-05-02T07:22:55",
"upload_time_iso_8601": "2024-05-02T07:22:55.473329Z",
"url": "https://files.pythonhosted.org/packages/01/9d/3ab344bcc6abe18e5d2964e2daa8edfaf1362814192e1b9fed015b11e08c/torch_model_manager-0.2.0-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "d8368336c07cf05cc134b5135cb7218b6d083d68321e82b592ca5d53497f236f",
"md5": "471d9143efc9c92682b03393a4881ac5",
"sha256": "a37c08c0ba59f369f97b7991fb73a60ed8bf892cc62ffcd9573a6e11a84ce3a9"
},
"downloads": -1,
"filename": "torch_model_manager-0.2.0.tar.gz",
"has_sig": false,
"md5_digest": "471d9143efc9c92682b03393a4881ac5",
"packagetype": "sdist",
"python_version": "source",
"requires_python": null,
"size": 16224,
"upload_time": "2024-05-02T07:22:57",
"upload_time_iso_8601": "2024-05-02T07:22:57.537940Z",
"url": "https://files.pythonhosted.org/packages/d8/36/8336c07cf05cc134b5135cb7218b6d083d68321e82b592ca5d53497f236f/torch_model_manager-0.2.0.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2024-05-02 07:22:57",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "Billal-MOKHTARI",
"github_project": "torch-model-manager",
"travis_ci": false,
"coveralls": false,
"github_actions": true,
"lcname": "torch-model-manager"
}
JSON
