#!/bin/bash
#SBATCH --job-name=p5_evaluate
#SBATCH --partition=akya-cuda
#SBATCH --nodes=1
#SBATCH --gres=gpu:4
#SBATCH --cpus-per-task=40
#SBATCH --mem=192G
#SBATCH --time=12:00:00
#SBATCH --output=/arf/scratch/stakan/hitit-proje/logs/p5/10_evaluate_%j.log
#SBATCH --error=/arf/scratch/stakan/hitit-proje/logs/p5/10_evaluate_%j.err

source /arf/scratch/stakan/hitit-proje/hitit_ocr/scripts/pipeline_p5/_common.sh

python3 hitit_ocr/src/evaluate_e2e.py \
  --detection-ckpts hitit_ocr/runs/p5/yolo_fold{0..4}/best.pt \
  --classifier-ckpts \
      hitit_ocr/runs/p5/hitit_dinov3_cls/best.pt \
      hitit_ocr/runs/p5/hitit_convnext_cls/best.pt \
      hitit_ocr/runs/p5/hitit_siglip_cls/best.pt \
  --seq2seq-ckpt hitit_ocr/runs/p5/seq2seq_image_ft/best.pt \
  --byt5-ckpt hitit_ocr/runs/p5/hitit_byt5/best.pt \
  --knn-store hitit_ocr/runs/p5/knn_store/ \
  --confusion-dir hitit_ocr/runs/p5/confusion_pairs/ \
  --moe-ckpt hitit_ocr/runs/p5/period_moe/best.pt \
  --gold-set datasets/eval_gold/manifest.jsonl \
  --output hitit_ocr/runs/p5/eval/