运行paddle_infer.py报错

[magicleo]

环境：
python 3.9
paddlepaddle-gpu
ubuntu 18.04
cuda11.7
tensorrt 8.5.3.1-1+cuda11.8 amd64 Meta package for TensorRT

1.修改models/rank/deepfm/config.yaml
use_inference: True

2.执行
python -u ../../../tools/static_trainer.py -m config.yaml
3. 执行
python -u ../../../tools/paddle_infer.py --model_file=output_model_deepfm/2/rec_inference.pdmodel --params_file=output_model_deepfm/2/rec_inference.pdiparams --use_gpu=True --data_dir=data/sample_data/train --reader_file=criteo_reader.py --batchsize=5 --enable_tensorRT=True

报错

E0118 19:50:24.366691 13690 helper.h:131] 4: [graphShapeAnalyzer.cpp::analyzeShapes::1872] Error Code 4: Miscellaneous (IShuffleLayer (Unnamed Layer* 10) [Shuffle]: reshape changes volume to multiple of original volume. Reshaping [13,9] to [1,1].) Traceback (most recent call last): File "/data/ymtguest/liuxinglong/ymt/PaddleRec/models/rank/deepfm/../../../tools/paddle_infer.py", line 188, in <module> main(args) File "/data/ymtguest/liuxinglong/ymt/PaddleRec/models/rank/deepfm/../../../tools/paddle_infer.py", line 126, in main predictor, pred_config = init_predictor(args) File "/data/ymtguest/liuxinglong/ymt/PaddleRec/models/rank/deepfm/../../../tools/paddle_infer.py", line 104, in init_predictor predictor = create_predictor(config) ValueError: (InvalidArgument) Errors occures in Paddle-TRT reshape2 op, try to use C++ Api config.Exp_DisableTensorRtOPs({"reshape2"}) ; or Python Api config.exp_disable_tensorrt_ops(["reshape2"]) to forbid reshape2 op into Paddle-TRT. [Hint: Expected layer->getOutput(0)->getDimensions().nbDims >= 0, but received layer->getOutput(0)->getDimensions().nbDims:-1 < 0:0.] (at ../paddle/fluid/inference/tensorrt/convert/reshape_op.cc:74)