The ALCF AI Testbed's Cerebras CS-2 system (consisting of two nodes) has been upgraded to a Cerebras Wafer-Scale Cluster, including the MemoryX system and SwarmX fabric, and is available to users.
Cerebras CS-2 is a wafer-scale, deep learning accelerator comprised of 850,000 processing cores, each providing 48KB of dedicated SRAM memory for an on-chip total of 40GB and interconnected to optimize bandwidth and latency. Its software platform integrates machine learning frameworks such as TensorFlow and PyTorch.
The ALCF's CS-2 system is designed to support large-scale models (up to and well beyond one billion parameters) and large-scale inputs.
Researchers can now submit project proposals for the Cerebras CS-2 system via the ALCF's Director's Discretionary program. There is no submission deadline; proposals will be accepted throughout the year.
For your convenience, here are links to the Cerebras CS-2 user guides and our allocation request form:
- ALCF Cerebras CS-2 user guide: https://docs.alcf.anl.gov/ai-testbed/cerebras/system-overview/
- Cerebras CS-2 user guide: https://docs.cerebras.net/en/latest/index.html
- Request an allocation: https://accounts.alcf.anl.gov/#/accountRequest
Please contact support@alcf.anl.gov with any questions.