mckinziebrandon/DeepChatModels

How to using the pre trained model to train my own dataset?

jackNhat opened this issue · 2 comments

First, thanks for great repository. I am newbie Chatbot, it helpful for me.
I have trained my own dataset and I have created my own model from scratch but it doesn't give good results.
I have added data then continue to train. But, I'm getting the following error:

ValueError: Cannot use 'decoder/while/decoder' as input to 'Merge/MergeSummary' because 'decoder/while/decoder' is in a while loop. See info log for more details.

How to solve the problem?

Thanks in advance!!!

Hello, thanks for checking out the repo. Unfortunately, since this repo is written in TF version 1.2, it is too old for me to maintain anymore. I'd use it more as a reference for creating your own architectures, but the code itself is rather outdated (TF2.0 looks much different).

With all that said, the error you are getting is saying that a tf.summary call was used somewhere on a tensor related-to or output-from a variable created by Decoder inside the body of a tf.while_loop. For a quick fix, you could remove any tf.summary calls related to the decoder code.

Thanks for reply @mckinziebrandon
I downgraded the tf version to 1.2 and it worked.
Once again, thank you.