Whenever I retrain the textsum model, it gives the following error:
InvalidArgumentError: Assign requires shapes of both tensors to match.
lhs shape= [35313] rhs shape= [35314]" exist.
I am using the same command line for training as provided on Github. I have set the number of GPUs to be used to 1. I have to train the already trained model because I have a large dataset which cannot be trained at once due to some time constraints. The model has made many checkpoints file when trained for the first time.