https://github.com/stephantul/torchic

Simple linear thing in Torch, with a scikit-learn compatible API.
https://github.com/stephantul/torchic

Last synced: about 6 hours ago
JSON representation

Simple linear thing in Torch, with a scikit-learn compatible API.

• Host: GitHub
• URL: https://github.com/stephantul/torchic
• Owner: stephantul
• License: mit
• Created: 2023-08-16T13:12:12.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2023-08-24T10:47:10.000Z (over 1 year ago)
• Last Synced: 2023-08-24T12:32:28.467Z (over 1 year ago)
• Language: Python
• Size: 102 KB
• Stars: 1
• Watchers: 3
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        


  

    

    

  

  


  




# torchic

Simple model training in Pytorch, with a scikit-learn compatible API.

It has the following features:

* Scikit-learn like API (i.e., using `fit` and `predict`)

* Supports numpy arrays and torch tensors out of the box

* Automatically converts your tensors between devices

## Example

The example below classifies 20 newsgroups, which is pre-vectorized using a CountVectorizer, courtesy of scikit-learn. This example requires that scikit-learn is installed.

```python

import numpy as np

from sklearn.datasets import fetch_20newsgroups_vectorized

from sklearn.model_selection import train_test_split

from sklearn.metrics import precision_recall_fscore_support

from torchic import Torchic

# NOTE: change this to 'cuda' or 'mps' if you want acceleration.

DEVICE = "cpu"

X, y = fetch_20newsgroups_vectorized(return_X_y=True, remove=("headers", "footers"), subset="train")

X = X[y < 10]

y = y[y < 10]

X = np.asarray(X.todense())

X_train, X_test, y_train, y_test = train_test_split(X, y, shuffle=True, random_state=44, test_size=.1)

n_features, n_labels = X_train.shape[1], len(set(y))

# Torchic stuff begins here.

t = Torchic(n_features, n_labels, learning_rate=1e-4).to(DEVICE)

t.fit(X_train, y_train, batch_size=128)

pred = t.predict(X_test)

print(precision_recall_fscore_support(y_test, pred, average="macro"))

```

## TODO:

* Add docstrings

* Add additional unit tests