Created by: sonixixi

复现步骤: 平台 tx2 , v2.3.0 源码编译

1. paddlehub ` import paddlehub as hub

module = hub.Module(name="pyramidbox_lite_mobile")

print(module.version)

module.processor.save_inference_model(dirname="pyramidbox_lite_mobile") `

2. 转换为lite格式 opt_mac --model_file=pyramidbox_lite_mobile/__model__ --param_file=pyramidbox_lite_mobile/__params__ --valid_targets=arm --optimize_out_type=naive_buffer --optimize_out=pyramidbox_lite_mobile

3. 使用paddle-lite 加载 ` // 1. Set MobileConfig, model_file_path is // the path to model model file. MobileConfig config; const std::string aa = "pyramidbox_lite_mobile.nb"; config.set_model_from_file(aa);

   // 2. Create PaddlePredictor by MobileConfig std::shared_ptr predictor = CreatePaddlePredictor(config); `

加载报 terminate called after throwing an instance of 'std::invalid_argument' what(): unmatched type, store as -746832279, but want to get N6paddle4lite10TensorLiteE