| #!/bin/bash | |
| #SBATCH --job-name=ssl_1node | |
| #SBATCH --partition=kolyoz-cuda | |
| #SBATCH --constraint=H100 | |
| #SBATCH --nodes=1 | |
| #SBATCH --gres=gpu:4 | |
| #SBATCH --cpus-per-task=64 | |
| #SBATCH --mem=128G | |
| #SBATCH --time=6:00:00 | |
| #SBATCH --output=/arf/scratch/stakan/hitit-proje/logs/h100/01_ssl_1node_%j.log | |
| #SBATCH --error=/arf/scratch/stakan/hitit-proje/logs/h100/01_ssl_1node_%j.err | |
| source /arf/scratch/stakan/hitit-proje/hitit_ocr/scripts/pipeline_h100/_common.sh | |
| torchrun --standalone --nproc_per_node=4 \ | |
| hitit_ocr/src/train_ssl_dinov3.py \ | |
| --config hitit_ocr/configs/ssl_dinov3_continual.yaml \ | |
| --output hitit_ocr/runs/h100/ssl_dinov3/ | |