football project, 我导出了lstm模型，在precdict时候报错，但用checkpoint里面的模型却能运行，不知道为啥

[freeman-1995]

Traceback (most recent call last): File "predict.py", line 28, in bmn_results, action_results = model_predict.infer(imgs_path, pcm_path) File "action_detect/action.py", line 37, in timer retval = func(*args) File "action_detect/action.py", line 104, in infer action_results = self.video_classify(material) File "action_detect/action.py", line 37, in timer retval = func(*args) File "action_detect/action.py", line 113, in video_classify action_results = self.classify_model.predict(self.configs, material=material) File "action_detect/models/lstm_infer.py", line 97, in predict output1, output2 = self.infer(input1_arr, input1_lod, input2_arr, input2_lod) File "action_detect/models/lstm_infer.py", line 65, in infer self.predictor.run() ValueError: In user code: InvalidArgumentError: The input tensor's dimension should be equal to the axis's size. But received input tensor's dimension is 2, axis's size is 3 [Hint: Expected x_rank == axis_size, but received x_rank:2 != axis_size:3.] (at /paddle/paddle/fluid/operators/transpose_op.cc:47) [operator < transpose2 > error]

[littlerookie]

一样的报错，请问您解决了吗？

[lxz-dream]

我也遇见了一样的问题，请问你们解决了吗

[jerrywgz]

您好，请问问题还是否需要解决，目前多模态视频能力集成在PaddleMIX中，https://github.com/PaddlePaddle/PaddleMIX/tree/develop 可以在这个repo下提出您的需求