Skip to content

Instantly share code, notes, and snippets.

View Spotlight0xff's full-sized avatar

André Merboldt Spotlight0xff

View GitHub Profile
@Spotlight0xff
Spotlight0xff / train-rnnt.slurm
Created March 3, 2021 21:12
Slurm file for training a Transducer model on 8 GPUs
#!/usr/local_rwth/bin/zsh
#SBATCH --mail-user=andre.merboldt@rwth-aachen.de
#SBATCH --mail-type=ALL
#SBATCH --nodes=1
#SBATCH --ntasks-per-node=8
#SBATCH --cpus-per-task=2
#SBATCH --time=120:00:00 # 5 days
#SBATCH --output=log/train-multi-gpu-job.%J.log # stdout/stderr file
#SBATCH --partition=dgx2 # DGX2 (has 16 GPUs)
#SBATCH --account=supp0003 # substitute appropriate group here