diff --git a/configs/kr_lp_pgnet.yml b/configs/kr_lp_pgnet.yml
index d8fcc80..77577be 100644
--- a/configs/kr_lp_pgnet.yml
+++ b/configs/kr_lp_pgnet.yml
@@ -1,6 +1,6 @@
 Global:
   use_gpu: True
-  epoch_num: 200
+  epoch_num: 50
   log_smooth_window: 20
   print_batch_step: 10
   save_model_dir: ./output/kr_lp_pgnet/
diff --git a/scripts/run_step1.sh b/scripts/run_step1.sh
new file mode 100755
index 0000000..7f8939c
--- /dev/null
+++ b/scripts/run_step1.sh
@@ -0,0 +1,49 @@
+#!/usr/bin/env bash
+# Step1 pretrain — 합성 데이터로 PGNet 학습
+#
+# 컨테이너 안 실행:
+#   docker exec kr_lp_pgnet bash /workspace/kr_lp_pgnet/scripts/run_step1.sh
+#
+# 환경 변수:
+#   DRY_RUN=1   2 epoch만 돌려 동작 검증
+#   EPOCHS=N    epoch 수 override (기본 config의 epoch_num)
+#   LOG=path    로그 파일 (기본: /workspace/PaddleOCR/output/kr_lp_pgnet/train.log)
+
+set -euo pipefail
+
+PADDLEOCR_DIR=/workspace/PaddleOCR
+KR_LP_DIR=/workspace/kr_lp_pgnet
+TRAIN_DATA=/workspace/train_data
+LOG="${LOG:-$PADDLEOCR_DIR/output/kr_lp_pgnet/train.log}"
+
+cd "$PADDLEOCR_DIR"
+
+# train_data symlink (config는 ./train_data/kr_lp_synth 사용)
+if [ ! -e ./train_data ]; then
+    ln -sf "$TRAIN_DATA" ./train_data
+fi
+
+mkdir -p "$(dirname "$LOG")"
+
+OVERRIDE=(
+    -o Global.pretrained_model=./pretrain_models/train_step1/best_accuracy
+       Global.load_static_weights=False
+)
+if [ -n "${EPOCHS:-}" ]; then
+    OVERRIDE+=(Global.epoch_num="$EPOCHS")
+fi
+if [ "${DRY_RUN:-0}" = "1" ]; then
+    OVERRIDE+=(Global.epoch_num=2 Global.eval_batch_step="[0,200]")
+    echo "DRY_RUN=1 → 2 epoch만 실행"
+fi
+
+echo "==========================="
+echo "Step1 학습 시작"
+echo "  config:  configs/e2e/kr_lp_pgnet.yml"
+echo "  data:    $TRAIN_DATA/kr_lp_synth/"
+echo "  pretrain: pretrain_models/train_step1/best_accuracy"
+echo "  log:     $LOG"
+echo "  override: ${OVERRIDE[@]}"
+echo "==========================="
+
+python3.10 tools/train.py -c configs/e2e/kr_lp_pgnet.yml "${OVERRIDE[@]}" 2>&1 | tee "$LOG"