Wgsl Python Packages

pyborch

PyTorch's API that trains in a browser tab on WebGPU — autograd, nn, optim, capture/replay, int8, ONNX out; held to real torch's values on 4,764 golden cases; measured on one page against TF.js, jax-js, Burn and ONNX Runtime Web (4–12× faster training, ahead of ORT at inference)

perturbation-kernel

Reproducible perturbation-kernel estimators. Scalar, SIMD and GPU backends that agree bit for bit, in Rust and Python.

ameva-forge

High-performance WebGPU-accelerated tensor computation engine with reverse-mode autograd for browser and edge deep learning.

moregpu-client

Native GPU compute pool: workers on any OS join an admin server with a one-liner (outbound WebSocket) and contribute GPU or CPU compute; jobs are sharded, AES-GCM sealed, computed, and pooled. Per-pool token wizard, TLS, throttling, self-healing services.

rocm-scribe

Automated CUDA-to-ROCm GPU kernel translation -- break free from CUDA lock-in with semantic BridgeIR

py2wgsl

Compile a restricted subset of Python into WGSL compute shaders that run on WebGPU

torch-webgpu

PyTorch compiler and WebGPU runtime