Free GPU resources

Google Colab

Each run can use a maximum of one GPU, which is usually an NVIDIA T4. The time limit for consecutive usage is 12 hours. The availability of the GPU resource is not guaranteed.

University of Leicester students/staff can use two HPCs for free: Alice and Sulis.

Alice

It has six GPU nodes. Each has 128GB RAM, 28 CPU cores and 2 x NVIDIA Tesla P100 GPU cards.

Access to Alice must be applied for, as described here. Access is organised in terms of projects, and you need to join a project to get access: (1) find out which project you can join by asking your colleagues or supervisor, then (2) ask your supervisor to email the IT desk requesting access on your behalf.

Documentation for Alice, e.g., its file system and hardware architecture, is available here.

Sulis

It has 30 GPU nodes. Each node has 3 NVIDIA A100 40GB GPUs (passively cooled) and 2 x AMD EPYC 7742 (Rome) 2.25GHz 64-core processors.

Access to Sulis can be applied for by following the instructions here, which also cover Sulis's documentation, e.g., file system and hardware specifications. When applying for a Sulis account, an SSH key is required, which can be generated via the Terminal (Linux/macOS) or Command Prompt (Windows). The SSH page also covers this. The username for your account should be your university username, e.g., ab123. Do not use PuTTY the first time you log in — it will skip the QR code, and you won't be able to scan it.

Details on how to use the HPCs for training ML models are available here.

The available resources on Sulis can be found here.