(日本語は英語の後に続きます。) A bit of exciting news to share today ✨ The HAMi community recently featured Midokura’s contribution to AMD support on their official account 🤗 Through this work, we can now see a future where NVIDIA and AMD accelerators can be managed through a single interface 👏 This is an important step toward Midokura’s vision of optimized AI infrastructure—enabling organizations to choose the right accelerator for the right workload and maximize efficiency, performance, and flexibility. A big thank you to the HAMi team for the recognition and great collaboration! ------------------------- 今日はちょっと嬉しいニュースです ✨ Midokuraが取り組んできたHAMiのAMDサポートについて、HAMi公式アカウントにて紹介されました 🤗 今回の取り組みで、NVIDIAとAMDがひとつのインターフェースで扱える未来が見えてきました 💪「ワークロードに合ったアクセラレータの選択」により、Midokuraの目指すAIインフラの「最適化」が現実のものに。 HAMiチームの皆さん、ご紹介をありがとうございます 🎵 #HAMi #AMD #OpenSource #Kubernetes #Midokura
Excited to see the first public exploration of AMD GPU virtualization with HAMi from Kenji Shimizu and the Midokura team. Their work demonstrates GPU sharing, memory isolation, and Compute Unit (CU) partitioning on AMD Instinct GPUs running on Kubernetes. It's a strong validation that HAMi's heterogeneous architecture can extend to new accelerator backends while keeping a consistent user experience. The work is already being discussed with the HAMi community through Issue #1707, opening the path toward upstream collaboration and future integration. Many thanks to Kenji Shimizu and the Midokura team for sharing both the implementation details and the technical discussions with the community. Open source grows when new ideas become shared infrastructure. We look forward to continuing the collaboration and exploring heterogeneous AI infrastructure together. Article: https://lnkd.in/gDq_RzTi Discussion: https://lnkd.in/giv6qVU3 #HAMi #Kubernetes #GPU #AMD #AIInfrastructure #OpenSource #CNCF