Metadata-Version: 2.1
Name: torch-fn
Version: 0.0.2
Summary: A decorator for seamless PyTorch calculations (primarily on CUDA) from numpy.ndarray and pd.DataFrame.
Home-page: https://github.com/ywatanabe1989/torch_fn
Author: ywatanabe1989
Author-email: ywata1989@gmail.com
License: MIT
Keywords: torch,GPU,numpy,pandas
Platform: UNKNOWN
Classifier: Programming Language :: Python :: 3
Classifier: License :: OSI Approved :: MIT License
Requires-Python: >=3.6
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: numpy
Requires-Dist: torch
Requires-Dist: pandas

![CI](https://github.com/ywatanabe1989/torch_fn/actions/workflows/pip_install.yml/badge.svg)

## Installation
``` bash
$ pip install torch_fn
```

## Usage

``` python
from torch_fn import torch_fn

import numpy as np
import torch.nn.functional as F

@torch_fn
def torch_softmax(*args, **kwargs):
    return F.softmax(*args, **kwargs)

def custom_print(x):
    print(type(x), x)

# Test the decorator with different input types
x = [1, 2, 3]
x_list = x
x_tensor = torch.tensor(x).float()
x_tensor_cuda = torch.tensor(x).float().cuda()
x_array = np.array(x)
x_df = pd.DataFrame({"col1": x})

custom_print(torch_softmax(x_list, dim=-1))
# /home/ywatanabe/proj/torch_fn/src/torch_fn/_torch_fn.py:57: UserWarning: Converted from  <class 'list'> to <class 'torch.Tensor'> (cuda:0)
#   warnings.warn(
# <class 'numpy.ndarray'> [0.09003057 0.24472848 0.6652409 ]

custom_print(torch_softmax(x_array, dim=-1))
# /home/ywatanabe/proj/torch_fn/src/torch_fn/_torch_fn.py:57: UserWarning: Converted from  <class 'numpy.ndarray'> to <class 'torch.Tensor'> (cuda:0)
#  warnings.warn(
# <class 'numpy.ndarray'> [0.09003057 0.24472848 0.6652409 ]

custom_print(torch_softmax(x_df, dim=-1))
# /home/ywatanabe/proj/torch_fn/src/torch_fn/_torch_fn.py:49: UserWarning: Converted from  <class 'pandas.core.frame.DataFrame'> to <class 'torch.Tensor'> (cuda:0)
#   warnings.warn(
# <class 'numpy.ndarray'> [0.09003057 0.24472848 0.6652409 ]

custom_print(torch_softmax(x_tensor, dim=-1))
# <class 'torch.Tensor'> tensor([0.0900, 0.2447, 0.6652])

custom_print(torch_softmax(x_tensor_cuda, dim=-1))
# <class 'torch.Tensor'> tensor([0.0900, 0.2447, 0.6652], device='cuda:0')
```


