| #!/bin/bash | |
| #SBATCH --job-name=p5_seq2seq_ft | |
| #SBATCH --partition=kolyoz-cuda | |
| #SBATCH --constraint=H100|H200 | |
| #SBATCH --nodes=1 | |
| #SBATCH --gres=gpu:4 | |
| #SBATCH --cpus-per-task=64 | |
| #SBATCH --mem=128G | |
| #SBATCH --time=4:00:00 | |
| #SBATCH --output=/arf/scratch/stakan/hitit-proje/logs/h100/08_seq2seq_ft_%j.log | |
| #SBATCH --error=/arf/scratch/stakan/hitit-proje/logs/h100/08_seq2seq_ft_%j.err | |
| source /arf/scratch/stakan/hitit-proje/hitit_ocr/scripts/pipeline_h100/_common.sh | |
| torchrun --standalone --nproc_per_node=4 \ | |
| hitit_ocr/src/seq2seq/train_image.py \ | |
| --config hitit_ocr/configs/seq2seq_hitit.yaml \ | |
| --phase finetune \ | |
| --encoder hitit_ocr/runs/h100/ssl_dinov3/checkpoint.pt \ | |
| --decoder hitit_ocr/runs/h100/seq2seq_text_pretrain/checkpoint \ | |
| --data datasets/processed/seq2seq_pairs.jsonl \ | |
| --output hitit_ocr/runs/h100/seq2seq_image_ft/ | |