The dimensions of the input data are always the same and there are no missing parts.

But why does an error occur when running two iterations well?

Error Message

File “”, line 101, in call

Determined shape must either match input shape along split_dim exactly if fully specified, or be less than the size of the input along split_dim if not fully specified. Got: 5120
[[{{node dmrl_2/split_7}}]] [Op:__inference_train_function_54827]

p.s. It is generated in Colab google.

I explored my data.

train_user shape: (2400,)
train_item shape: (2400,)
train_text shape: (2400, 768)
train_image shape: (2400, 4096)
train_y.shape: (2400,)

But, There is no missing data.