Translate embedding to natural language

[peasypi]

Hello,
I have trained a model that returns a clip embedding (so far at least a tensor of the right size). I now want to try to convert this tensor into natural language and came across this repo. To use it (as written in the readme), I have to provide an image. But now I only have the embedding. Is it possible to use your code only to "translate" an embedding into a natural language sentence?
Thank you in advance!