I create the training set as a window of 11 size and stride 1 to the end of the corpus like this
For example my training set x shape is 512,11 (512 is bs and 11 is inputs) but when I predict the input shape is 1,20. Is it affect the performance of the model ?. I think that I trained the model with fix-length so in the predicting process I should do the same as training.
English is not my mother tongue if something is not clear pls tell me, Thank you