Hosted at ULB, the Lyra cluster is a hyperconverged virtual HPC infrastructure designed for Machine Learning (ML), Artificial Intelligence (AI), and High Throughput Computing (HTC) workloads. Each virtual node has one NVIDIA RTX 6000 ADA (48 GB) and 32 cores EPYC-Genoa with 128 GB RAM.

Suitable for:

ML/AI and high-throughput workloads (CUDA, TensorFlow, PyTorch, etc.); 5 days Wall Time.

Resources

  • Home directory (100 GB per user)
  • Global working directory /globalsc ($GLOBALSCRATCH) — 5 TB per user
  • Node local working directory $LOCALSCRATCH (defined dynamically in jobs)
  • Default batch queue*

Access / Support

SSH: lyra.ulb.be (port 22) via your university gateway, with proper login and id_rsa.ceci file.

Support: ceci-support@ulb.be

Server SSH key fingerprints:

What’s this ?

ECDSA:

SHA256:qND98sfbhXnPC6vkS/r/whUpIjPoSiVCuxAojG6N2jI

ED25519:

SHA256:CGQINW7XKq2ZKuvJBvGrmQJG5E+pjBILmeECbiXZ0Ug

RSA:

SHA256:zGzVwKSR0AZaoyX43Mmi95byqxI3NmSzysL66E0KA+c

cluster image 1
cluster image 2