Closed colebuckleyy closed 2 years ago
Unfortunately no :-)
The closest thing I can think about is encoding the entire training set, and saving a database of <example,vector> pairs.
Then at test time, you can encode the test example, and using the test code vector - look for the nearest neighbor training vector from the database.
Ok, thank you
I'm wondering if its somehow possible to convert a code vector back into its text format in contexts, is this possible?