PaddleOCR-VL LoRA 微调后 怎么合并模型？ #1405
Description
erniekit export script/conf/run_ocr_vl_sft_16k_lora.yaml lora=True
报错：
LAUNCH INFO 2025-12-20 02:34:18,837 ------------------------- ERROR LOG DETAIL -------------------------
ation file /kanas/public/zhangfa/sft_paddle/outputs/official_models/PaddleOCR-VL/config.json
Traceback (most recent call last):
File "/home/ERNIE/erniekit/launcher.py", line 58, in
launch()
File "/home/ERNIE/erniekit/launcher.py", line 48, in launch
run_export()
File "/home/ERNIE/erniekit/export/export.py", line 171, in run_export
mergekit.merge_model()
File "/usr/local/lib/python3.10/dist-packages/paddleformers/mergekit/merge_model.py", line 94, in merge_model
self.merge_lora_model()
File "/usr/local/lib/python3.10/dist-packages/paddleformers/mergekit/merge_model.py", line 570, in merge_lora_model
self.merge_safetensor_lora_model(file_type_list)
File "/usr/local/lib/python3.10/dist-packages/paddleformers/mergekit/merge_model.py", line 648, in merge_safetensor_lora_model
import_class = importlib.import_module(f"paddleformers.transformers.{config_dict['model_type']}.modeling")
File "/usr/lib/python3.10/importlib/init.py", line 126, in import_module
return _bootstrap._gcd_import(name[level:], package, level)
File "", line 1050, in _gcd_import
File "", line 1027, in _find_and_load
File "", line 992, in _find_and_load_unlocked