Ember

Ember is a lightweight deep learning Python package with C and CUDA implementations, designed for performance and simplicity.

Features

• Lightweight: Minimal overhead, focused on core tensor operations.
• Hardware Accelerated:
• Pure C implementation for CPU.
• CUDA implementation for NVIDIA GPUs.
• NumPy Compatible: Seamless conversion to and from NumPy arrays.
• Autograd Support: (Upcoming) Automatic differentiation for training neural networks.

Quick Start

import ember
import numpy as np

# Create a tensor from a list
t = ember.Tensor([1.0, 2.0, 3.0])

# Perform operations
t2 = ember.sin(t)
print(t2.to_list())
# Output: [0.84147096, 0.90929741, 0.14112000] (approx)