#!/usr/bin/env bash
# v23 training — Fix Track 4 concept grounding failures from v22.
#
# v22 eval results (KS=0.1527, Pearson=0.4433) — Track 2 PASS.
# v22 Track 4 failures:
#
#   blur → continuity   XX+0.261  (should be negative)
#   jpeg → continuity   XX+0.184  (should be negative)
#   occlusion → minutiae_reliability  XX+0.067  (should be negative)
#   dry_skin → contrast_uniformity   XX+0.504  (should be negative)
#
# Root cause: --concept-deg-gamma 0.5 is too weak.
#   Gaussian blur naturally SMOOTHS ridges → model's continuity feature increases
#   with blur. L_deg with gamma=0.5 cannot overcome this natural correlation.
#   contrast_uniformity is only targeted by dry_skin once → single weak signal.
#   minutiae_reliability for occlusion similarly has a weak signal.
#
# v23 fix: --concept-deg-gamma 2.0 (vs v22's 0.5 — 4× stronger concept supervision)
#   All other settings are identical to v22.
#
# Note: orientation_coherence saturation (mean=0.952, always near 1.0) is NOT
#   fixed in v23. It requires adding concept[0] to dry_skin and wet_press targets
#   in DEGRADATION_CONCEPT_MAP (T39 fix, applied to src/losses/degradation_ranking.py).
#   v24 will include this fix.

set -euo pipefail

REPO_ROOT="$(cd "$(dirname "$0")/../.." && pwd)"
export PATH="/home/aiserver/miniconda3/bin:$PATH"

VERSION="v23"
SAVE_DIR="${REPO_ROOT}/sifq/checkpoints/${VERSION}"
LOG_FILE="${REPO_ROOT}/sifq/logs/train_${VERSION}.log"
EVAL_SCRIPT="${REPO_ROOT}/sifq/scripts/run_eval_${VERSION}.sh"

mkdir -p "${REPO_ROOT}/sifq/logs"

python "${REPO_ROOT}/sifq/scripts/train_sifq.py" \
  --root-302a  "${REPO_ROOT}/dataset/302a/images/challengers" \
  --root-302b  "${REPO_ROOT}/dataset/302b/images/baseline" \
  --root-302d  "${REPO_ROOT}/dataset/nist_302d/images/auxiliary" \
  --root-fvc2002 "${REPO_ROOT}/dataset/FVC_Dataset/FVC2002" \
  --root-fvc2004 "${REPO_ROOT}/dataset/FVC_Dataset/FVC2004" \
  --root-polyu "${REPO_ROOT}/dataset/PolyU" \
  --exclude-sensor "R_1000_slap,R_500_slap,S_500_slap" \
  --mdgt-checkpoint "${REPO_ROOT}/pad/TRAM-downstream/checkpoint/checkpoints_dinov2_tram/best_eer.pt" \
  --epochs 60 \
  --batch-size 96 \
  --image-size 224 \
  --lr 1e-4 \
  --spread-mode uniform \
  --spread-weight 3.0 \
  --concept-deg-gamma 2.0 \
  --sd302-concept-weight 0.0 \
  --deg-every-n-steps 2 \
  --no-mat-stats \
  --proto-max-batches 0 \
  --k-cross 0 \
  --max-train-samples -1 \
  --num-workers 8 \
  --gpus 0 \
  --save-dir "${SAVE_DIR}"

echo "[auto-eval] Training done. Starting eval ${VERSION}..."
bash "${EVAL_SCRIPT}" >> "${LOG_FILE}" 2>&1