taoqi98 / FedNewsRec

Data and code of our paper "Privacy-Preserving News Recommendation Model Learning"
MIT License
39 stars 14 forks source link

ValueError: Dimension must be 5 but is 4 for #12

Open qwqwqwqw1 opened 6 months ago

qwqwqwqw1 commented 6 months ago

Hello. I got a question:

ValueError: Dimension must be 5 but is 4 for '{{node attention/transpose_7}} = Transpose[T=DT_FLOAT, Tperm=DT_INT32](attention/truediv, attention/transpose_7/perm)' with input shapes: [?,20,28,20,28], [4].

Call arguments received by layer "attention" (type Attention): • x=['tf.Tensor(shape=(None, 28, 400), dtype=float32)', 'tf.Tensor(shape=(None, 28, 400), dtype=float32)', 'tf.Tensor(shape=(None, 28, 400), dtype=float32)']

I did not know how the problem happens.I did not modify the code. Another thing is that I cannot follow the two specific versions tensorflow-gpu == 1.13.0 keras == 2.2.4. Maybe it is because they are two old versions? I am not sure whether the different versions cause this problem. If so, how should I solve it? Could you please give some suggestions? Looking forward to your reply.