pycuts 1.0.0

A small collection of Python shortcut scripts for common tasks.

pycuts

pycuts is a small Python library that provides a collection of shortcut functions for common operations across various libraries, particularly for PyTorch and Hugging Face Hub environments.

Installation

You can install pycuts directly from PyPI:

pip install pycuts

Shortcuts

Function	Return Type	Description
device()	torch.device	Determines the best device to use (cuda, mps, or cpu).
gpu()	bool	Returns True if a GPU ("cuda" or "mps") is available, otherwise False.
torch_dtype()	torch.dtype	Determines the appropriate tensor precision based on the device.
synchronize()	None	Waits for all kernels in all streams on the given device to complete.
empty_cache()	None	Clears the GPU memory to prevent out-of-memory errors.
device_count()	int	Returns the number of available devices (e.g., number of GPUs).
manual_seed(seed)	None	Sets the random seed for reproducible behavior across CPU/GPU.
is_spaces()	bool	Returns True if running in a Hugging Face Space, otherwise False.
is_zero_gpu_space()	bool	Returns True if running in a zero-GPU Hugging Face Space.

Examples

import pycuts
print(f"current device is: {pycuts.device()}")