Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/antoinebrl/torchextractor
Feature extraction made simple with torchextractor
https://github.com/antoinebrl/torchextractor
feature-extraction machine-learning python3 pytorch
Last synced: about 1 month ago
JSON representation
Feature extraction made simple with torchextractor
- Host: GitHub
- URL: https://github.com/antoinebrl/torchextractor
- Owner: antoinebrl
- License: apache-2.0
- Created: 2021-03-03T21:01:16.000Z (over 3 years ago)
- Default Branch: main
- Last Pushed: 2021-03-10T09:51:56.000Z (over 3 years ago)
- Last Synced: 2024-04-27T06:53:58.224Z (7 months ago)
- Topics: feature-extraction, machine-learning, python3, pytorch
- Language: Python
- Homepage: https://torchextractor.readthedocs.io
- Size: 31.3 KB
- Stars: 99
- Watchers: 2
- Forks: 4
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# `torchextractor`: PyTorch Intermediate Feature Extraction
[](https://pypi.org/project/torchextractor/)
[](https://pypi.org/project/torchextractor/)
[](https://torchextractor.readthedocs.io/en/latest/)
[](https://github.com/antoinebrl/torchextractor/actions/workflows/publish.yml)
[](https://github.com/antoinebrl/torchextractor/blob/main/LICENSE)
## Introduction
Too many times some model definitions get remorselessly copy-pasted just because the
`forward` function does not return what the person expects. You provide module names
and `torchextractor` takes care of the extraction for you.It's never been easier to
extract feature, add an extra loss or plug another head to a network.
Ler us know what amazing things you build with `torchextractor`!
## Installation
```shell
pip install torchextractor # stable
pip install git+https://github.com/antoinebrl/torchextractor.git # latest
```
Requirements:
- Python >= 3.6+
- torch >= 1.4.0
## Usage
```python
import torch
import torchvision
import torchextractor as tx
model = torchvision.models.resnet18(pretrained=True)
model = tx.Extractor(model, ["layer1", "layer2", "layer3", "layer4"])
dummy_input = torch.rand(7, 3, 224, 224)
model_output, features = model(dummy_input)
feature_shapes = {name: f.shape for name, f in features.items()}
print(feature_shapes)
# {
# 'layer1': torch.Size([1, 64, 56, 56]),
# 'layer2': torch.Size([1, 128, 28, 28]),
# 'layer3': torch.Size([1, 256, 14, 14]),
# 'layer4': torch.Size([1, 512, 7, 7]),
# }
```
[See more examples](docs/source/examples.ipynb)
[](https://mybinder.org/v2/gh/antoinebrl/torchextractor/HEAD?filepath=docs/source/examples.ipynb)
[](https://colab.research.google.com/github/antoinebrl/torchextractor/blob/master/docs/source/examples.ipynb)
[Read the documentation](https://torchextractor.readthedocs.io/en/latest/)
## FAQ
**• How do I know the names of the modules?**
You can print all module names like this:
```python
tx.list_module_names(model)
# OR
for name, module in model.named_modules():
print(name)
```
**• Why do some operations not get listed?**
It is not possible to add hooks if operations are not defined as modules.
Therefore, `F.relu` cannot be captured but `nn.Relu()` can.
**• How can I avoid listing all relevant modules?**
You can specify a custom filtering function to hook the relevant modules:
```python
# Hook everything !
module_filter_fn = lambda module, name: True
# Capture of all modules inside first layer
module_filter_fn = lambda module, name: name.startswith("layer1")
# Focus on all convolutions
module_filter_fn = lambda module, name: isinstance(module, torch.nn.Conv2d)
model = tx.Extractor(model, module_filter_fn=module_filter_fn)
```
**• Is it compatible with ONNX?**
`tx.Extractor` is compatible with ONNX! This means you can also access intermediate features maps after the export.
Pro-tip: name the output nodes by using `output_names` when calling `torch.onnx.export`.
**• Is it compatible with TorchScript?**
Not yet, but we are working on it. Compiling registered hook of a module
[was just recently added in PyTorch v1.8.0](https://github.com/pytorch/pytorch/pull/49544).
**• "One more thing!" :wink:**
By default we capture the latest output of the relevant modules,
but you can specify your own custom operations.
For example, to accumulate features over 10 forward passes you
can do the following:
```python
import torch
import torchvision
import torchextractor as tx
model = torchvision.models.resnet18(pretrained=True)
def capture_fn(module, input, output, module_name, feature_maps):
if module_name not in feature_maps:
feature_maps[module_name] = []
feature_maps[module_name].append(output)
extractor = tx.Extractor(model, ["layer3", "layer4"], capture_fn=capture_fn)
for i in range(20):
for i in range(10):
x = torch.rand(7, 3, 224, 224)
model(x)
feature_maps = extractor.collect()
# Do your stuffs here
# Discard collected elements
extractor.clear_placeholder()
```
## Contributing
All feedbacks and contributions are welcomed. Feel free to report an issue or to create a pull request!
If you want to get hands-on:
1. (Fork and) clone the repo.
2. Create a virtual environment: `virtualenv -p python3 .venv && source .venv/bin/activate`
2. Install dependencies: `pip install -r requirements.txt && pip install -r requirements-dev.txt`
4. Hook auto-formatting tools: `pre-commit install`
5. Hack as much as you want!
6. Run tests: `python -m unittest discover -vs ./tests/`
7. Share your work and create a pull request.
To Build documentation:
```shell
cd docs
pip install requirements.txt
make html
```