Green Computing with Remote GPU over APN (tsuzumi-7B)
In accordance with the PoC Reference for Green Computing with Remote GPU over APN, we conducted a proof of concept using the small‑scale LLM tsuzumi‑7B.For compute‑dominant workloads where the dataset is relatively small, data transmission time is negligible compared to GPU computation time, even in long‑distance remote GPU environments. While replication and cache strategies are […]
Green Computing with Remote GPU for AI Training with Image Data
In this report, we evaluate the feasibility of remote storage for AI training by connecting two sites via APN and using image data for Unet-3D model training to assess storage access performance. Through this, we aim to verify the effectiveness of APN in maintaining practical training time in a distributed environment.
