training problem

[xueyaosmile]

I used the repair data set I made for training. After the first epoch, the loss and loss remained unchanged. In addition, the loss kept increasing during the training process. What is the problem?

[HighwayWu]

Please double-check that the training data (or the generation of npy file) is correctly loaded.

[xueyaosmile]

请仔细检查训练数据（或 npy 文件的生成）是否正确加载。

Hello, I checked the data format, as shown below, there should be no problem, but the loss still increases during training, and the acc becomes 50

[HighwayWu]

Could you please try the training by using our GC_Places/Dresden data? And if it works, the problem may be your dataset is not suitable for this Xception-based network.

[xueyaosmile]

Could you please try the training by using our GC_Places/Dresden data? And if it works, the problem may be your dataset is not suitable for this Xception-based network.

Hello, I am training on the places dataset given by the author.