this is based off tensorflow's seq2seq code, but with slight modifications and generalized so any dataset can be applied to it
Given a folder that contains a 2 files, first.txt and last.txt with an equal number of lines, build a seq2seq model. This could be a translation model, email to response model, parser model, or any other text that can be modeled as: input-> output Both text files MUST contain the same amount of lines, each line corresponds to the other file's same line position
This code is meant to run on *nix/mac, not windows
to train:
source ~/tensorflow/bin/activate
python main.py --data_dir ~/data_sets/eng-fren/ --train_dir ~/train_dir
to inference:
source ~/tensorflow/bin/activate
python main.py --train_dir ~/train_dir --decode True --input "hello world"
the data dir assumes that there are 2 txt files.