scripts/archive/run_eval_v13.sh

#!/usr/bin/env bash
# Eval v13 (chính xác): loại non-segmented slap trước khi tính metrics.
#
# Lý do: R_1000_slap, R_500_slap, S_500_slap có std ≈ 0 (score ~21.2)
# vì ảnh full-hand slap visually homogeneous. Các sensors này kéo mean_KS
# lên 0.634 (xấu) và méo distribution. Eval đúng phải exclude chúng.
#
# Usage:
#   bash sifq/scripts/run_eval_v13.sh [training_pid]

set -euo pipefail

REPO_ROOT="/home/aiserver/works/fingerprint"
VENV="$REPO_ROOT/.venv"
CKPT="$REPO_ROOT/sifq/checkpoints_full_v13/last.pt"
SCORES="$REPO_ROOT/sifq/eval_results/sifq_scores_v13.jsonl"
OUTDIR="$REPO_ROOT/sifq/eval_results/v13_seg"
EXCLUDE="R_1000_slap,R_500_slap,S_500_slap"

source "$VENV/bin/activate"
cd "$REPO_ROOT"

TRAIN_PID="${1:-}"
if [[ -n "$TRAIN_PID" ]]; then
    echo "[eval_v13] Waiting for training PID $TRAIN_PID to finish..."
    while kill -0 "$TRAIN_PID" 2>/dev/null; do
        sleep 30
    done
    echo "[eval_v13] Training finished. Starting eval..."
fi

mkdir -p "$OUTDIR"

# ── Inference (chỉ chạy nếu SCORES chưa tồn tại) ─────────────────────────
if [[ ! -f "$SCORES" ]]; then
    echo "[eval_v13] Running inference on $CKPT ..."
    python sifq/scripts/run_infer.py \
        --checkpoint "$CKPT" \
        --exclude-sensor "$EXCLUDE" \
        --output "$SCORES"
    echo "[eval_v13] Inference done: $SCORES"
else
    echo "[eval_v13] Scores file already exists: $SCORES (skipping inference)"
fi

# ── Evaluation (exclude non-segmented slap) ────────────────────────────────
echo "[eval_v13] Running evaluation (Track 2 + Track 4, exclude: $EXCLUDE)..."
python sifq/scripts/run_eval.py \
    --sifq-scores "$SCORES" \
    --checkpoint  "$CKPT" \
    --out-dir     "$OUTDIR" \
    --exclude-sensor "$EXCLUDE" \
    --skip-track1

echo "[eval_v13] Eval done. Results in $OUTDIR"
echo ""
echo "=== SUMMARY ==="
cat "$OUTDIR/eval_summary.json"