Empowering the open-source ecosystem to plug in diverse accelerators

We aim to grow alongside the rapidly evolving open-source AI software ecosystem. We strive to realize our vision of making AI more accessible.

SGLang X Moreh

We enhance the performance of SGLang, a fast and efficient open-source serving engine for LLMs and VLMs, on AMD Instinct MI300 series GPUs by contributing our library-level and model-level optimizations.

TT-Metalium X Moreh

As one of the most active external contributors to TT-Metalium (see our pull requests and issues we submitted), we enhance its functionality, stability, and performance across both runtime system and kernels, while ensuring seamless integration with our inference and training frameworks.

AMD ROCm X Moreh

Through collaboration with AMD engineering team, we contributed to the MIOpen library (see our pull requests) by supporting various deep learning primitive operations and optimizing them for the AMD CDNA architecture.

SkyPilot X Moreh

We contribute to SkyPilot to improve support for AMD GPU clusters. This allows our AMD GPU customers to have an open-source-based platform option.