#tensor #blas #machine-learning

yanked ebcandle-core

Minimalist ML framework

2 releases

0.7.3 Oct 27, 2024
0.7.2 Oct 27, 2024

#26 in #blas

Download history 200/week @ 2024-10-25 51/week @ 2024-11-01 6/week @ 2024-11-08

70 downloads per month

MIT/Apache

1.5MB
34K SLoC

Rust 29K SLoC // 0.0% comments Metal Shading Language 2.5K SLoC // 0.0% comments CUDA 1.5K SLoC // 0.0% comments

Contains (Zip file, 2KB) tests/fortran_tensor_3d.pth, (Zip file, 2KB) tests/test.pt, (Zip file, 2KB) tests/test_with_key.pt

ebcandle

Minimalist ML framework for Rust


lib.rs:

ML framework for Rust

use ebcandle_core::{Tensor, DType, Device};

let a = Tensor::arange(0f32, 6f32, &Device::Cpu)?.reshape((2, 3))?;
let b = Tensor::arange(0f32, 12f32, &Device::Cpu)?.reshape((3, 4))?;

let c = a.matmul(&b)?;

Features

  • Simple syntax (looks and feels like PyTorch)
  • CPU and Cuda backends (and M1 support)
  • Enable serverless (CPU) small and fast deployments
  • Model training
  • Distributed computing (NCCL).
  • Models out of the box (Llama, Whisper, Falcon, ...)

FAQ

  • Why Candle?

Candle stems from the need to reduce binary size in order to enable serverless possible by making the whole engine smaller than PyTorch very large library volume

And simply removing Python from production workloads. Python can really add overhead in more complex workflows and the GIL is a notorious source of headaches.

Rust is cool, and a lot of the HF ecosystem already has Rust crates safetensors and tokenizers

Dependencies

~9–21MB
~354K SLoC