| --- |
| title: ROCmPort AI |
| emoji: ⚡ |
| colorFrom: red |
| colorTo: yellow |
| sdk: gradio |
| pinned: true |
| license: apache-2.0 |
| thumbnail: >- |
| https://cdn-uploads.huggingface.co/production/uploads/69cbcab7f7b604127917f905/meymza-6XkpBbcJO2LUR7.png |
| short_description: Agentic CUDA-to-ROCm migration loop |
| sdk_version: 6.14.0 |
| --- |
| |
|
|
| # ROCmPort AI |
|
|
| **Demo**: [View Live Demo](https://www.youtube.com/watch?v=adfGDrPrqcQ) |
|
|
| ## What it does |
|
|
| ROCmPort AI automatically migrates CUDA GPU code to ROCm (AMD's open-source GPU computing platform), enabling seamless portability across different GPU architectures. |
|
|
| ## Key Features |
|
|
| - 🚀 **Automated Code Translation** - Converts CUDA kernels and libraries to ROCm HIP code with minimal manual intervention |
| - 📊 **Performance Analysis** - Generates detailed migration reports with benchmark comparisons and optimization recommendations |
| - 🔧 **Smart Patching** - Intelligently handles library replacements, API mappings, and architecture-specific optimizations |
|
|
| --- |
|
|
| For detailed documentation and examples, see [BENCHMARKS.md](BENCHMARKS.md) and the [docs](docs/) folder. |
