Getting started¶
Installation¶
Install from source in editable mode:
pip install -e .
PyTorch 2.0 or newer and Python 3.12+ are required.
Quick example¶
Tensor math helpers:
import torch
from torchgear import abs2, expi
z = torch.tensor([1 + 2j, 3 + 4j])
abs2(z)
x = torch.tensor([0.0, torch.pi / 2])
expi(x)
Module base class with device/dtype helpers and freezing:
from torchgear.module import TorchgearModule
import torch
import torch.nn as nn
class MyModel(TorchgearModule):
def __init__(self):
super().__init__()
self.w = nn.Parameter(torch.randn(4))
def forward(self, x):
return x @ self.w
model = MyModel()
model.freeze()
y = model.new_zeros(2) # same device and dtype as parameters