old training model load error but new training is ok for colab

[jinxsfe]

Dear advisor , I ask for help for load my old training model with tensorflow version

and I want use the(load) model predict the data in recently but failed, the error showed below

, but I training again for today in same data, parameter, seems prediction is worked, the new model is

,

does any one can give me the ideas for how to restore old oversion, because I don't want to loose that model, I try to load the same GPU T4 or different GPU L4, it not working

[dhantule]

Hi @jinxsfe, thanks for reporting this.
Can you provide some code to reproduce ?

[jinxsfe]

https://colab.research.google.com/drive/1z3qcuJnF8MfPwC0U9BJAmhE-pDKH7qbg?usp=drive_link,

[jinxsfe]

@dhantule https://drive.google.com/drive/folders/1-p8YnT_4Vm6oF-f3ljrxlhUM-jLeTgxa?usp=drive_link,https://drive.google.com/drive/folders/1-UJBHa8mzP_ale89fiAmgZpHibpxVyQQ?usp=drive_link, I shared one old model and test data and you can check it,

[jinxsfe]

when using old model, direct skip 5.1,5.2,5.3,5.4

[dhantule]

Hi @jinxsfe, the links are not accessible to me. Can you provide access ?

[jinxsfe]

I can not see the access request in your mail, can you tell me your gmail and I can put permission an you can directly check it, @dhantule

[dhantule]

Hi @jinxsfe, just share a gist.

go to file -> save a copy as github gist , then share that gist here.