keras/keras/engine/training.py
Line 783 in 3a33d53
data = data_adapter.expand_1d(data)
There is a expand operation for 1d input data, but this op is missed in v2.7.0. I’d like to know why and how can i keep compatibility
Previously Model.fit/evaluate/predict will uprank 1d data to 2d, E.g. shape (batch,) to (batch, 1). This is potentially a breaking change for subclassed models that process scalar data as well as for certain sequential models that have been called with incorrectly shaped data. Now, only data of shape (batch_size, dim) would be accepted (i.e. rank 2). If you were in this situation, you need to update your Sequential input shape so that it has the same rank as the data you are going to pass to the model. Thank you