• Stands for vectorized N-dimensional numerical arrays. Tensor / Nd Array library.

• Currently capable of creating CPU Tensors of type T and performing

  • broadcasted algebraic operations
  • Nd matrix multiplication (naive)
  • 1d and 2d convolution / cross-correlation (naive) with strides
  • reduce operations such as sum, product, max, min and pooling
  • transformations such as view/reshape, permute/transpose, flip, expand, pad, slice, squeeze, unsqueeze

• Clone the repo and run examples

cargo run -r --example <example_name>

• Use as library

cargo add --git https://github.com/shettysach/veNum

credits

• kurtschelfthout/tensorken
• huggingface/candle
• minitorch/minitorch
• nreHieW/r-nn
• assets/venom.png

resources

• pytorch-internals by ezyang
• tensorken articles by Kurt Schelfthout
• MiniTorch
• Convolutions