Requirements:
Keras
TensorFlow or Theano
OpenNMT
GPU for fast execution
For running Basic encoder_decoder just run "python enc_dec.py"
For running models using openNMT go to the openNMT folder and put parllel corpus into data folder and run the following 3 commands
1) This command is for pre-processing
th preprocess.lua -train_src data/src-train.txt -train_tgt data/tgt-train.txt -valid_src data/src-val.txt -valid_tgt data/tgt-val.txt -save_data data/demo
2)This command is for training this varies with the configuration of your model you want to build.
th train.lua -data data/demo-train.t7 -save_model demo-model -gpuid 1 -brnn -brnn_merge sum -residual 1 -global_attention general -end_epoch 30
3)This is for translating
th translate.lua -model demo-model_epoch30_260.34.t7 -src data/src-test.txt -output pred.txt -gpuid 1
Model Evaluation:
Bleu:
perl bleu.pl hin_test.txt < pred.txt