Launching training job: transformer_train_medium
Nodes: 1, GPUs per node: 4
Resources per node: 32 CPUs, 128 GB memory
MASTER_ADDR: dcc-core-55
MASTER_PORT: 29689
SLURM_NNODES: 1
SLURM_NTASKS_PER_NODE: 
SLURM_PROCID: 0
CUDA_VISIBLE_DEVICES: 0,1
slurmstepd: error: *** JOB 42254198 ON dcc-core-55 CANCELLED AT 2026-01-20T20:55:48 ***