Paragon v2.0.0 – WebGPU Native + Browser AI Inference

Paragon v2.0.0 introduces full WebGPU acceleration for both native and browser environments

🚀 Highlights

• ✅ Native GPU support via WebGPU (Vulkan, Metal, DirectX12)
• ✅ In-browser neural network inference via WebAssembly + WebGPU
• ✅ Procedural WGSL shader generation (activations + types)
• ✅ Supports float32, int32, uint32 in both CPU and GPU modes
• ✅ Full WASM bindings via Go reflection for JavaScript

🔧 Refactoring & Structural Changes
🧱 WebGPU now lives in its own repo:
We've extracted and open-sourced the internal WebGPU bridge to openfluke/webgpu — a modular, zero-CGo Go wrapper around the official GPU Web API, supporting both native and browser targets.

🌀 Internal GPU logic refactored:
webgpu.go and webgpu_optimized.go have been restructured for clarity, modularity, and support of GPU compute pipelines that dynamically adapt to layer size and numeric type.

🔁 Dynamic shader generation:
Shader code (WGSL) is now built per-layer based on activation function and tensor type, allowing flexible GPU dispatch with no hardcoded logic.

🧠 Network API upgraded:
All public methods on the Network struct are now auto-exposed to JavaScript via reflect + WASM bindings. JS apps can create, run, and benchmark Paragon networks from the browser console or UI.