Serving: version names issue: No versions of servable found under base path

[EmnamoR]

Hi,

I am trying to build a production environment for our company where we can export keras/pythorch, etc models to savedmodels and deploy them with tensorflow serving.
I faced a problem of No versions of servable testv found under base path s3://

when i check s3 i can find the version. then i tried to export the same model with a diffirent version name which is 1 and it is working.

so any version name like 0.1 or v0.1 wont work. Does anybody has any information about this. How should we name our versions.

Thank you very much

[Harshini-Gadige]

Duplicate of #266
A similar issue is addressed in StackOverflow and here is the link for the same. Please note that any support related questions are to be posted in StackOverflow as we encourage users to post a bug or a feature request here. Thanks !

[linyia01]

How did you solve it？

[EmnamoR]

@linyia01
as far as i remember versions should be on a specific format like :
0000001 and 0000002
i kept the versions in an incremental way.
i hope this will help

[sngjuk]

@linyia01 change directory structure, and designate model directory as same (my_models/model)

my_models/model [assets,saved_model.pb,variables] << change it to add version number
my_models/model/1 [assets,saved_model.pb,variables]

stack overflow