@ts-torch/core

Core tensor operations & FFI bindings for ts-torch.

Overview

This package provides the foundational tensor operations and Foreign Function Interface (FFI) bindings for ts-torch, enabling high-performance tensor computations backed by native implementations.

Features

• Tensor Operations: Core tensor class and operations (creation, manipulation, etc.)
• FFI Bindings: Low-level bindings to native tensor operations and PyTorch C++ API
• Type Definitions: Comprehensive TypeScript type definitions for tensor operations
• Memory Management: Efficient memory pooling and tracking for tensor allocations

Installation

bun add @ts-torch/core

Usage

import { tensor, zeros, ones } from '@ts-torch/core'

// Create a tensor from data
const t = tensor([
  [1, 2, 3],
  [4, 5, 6],
])
console.log(t.shape) // [2, 3]

// Create zero/one tensors
const z = zeros([3, 3])
const o = ones([2, 4])

Development

# Build the package
bun run build

# Watch mode
bun run dev

# Type checking
bun run check-types