MemoryAugDialog

This is the accompany python package for the ACL 2019 paper：Learning to Abstract for Memory-augmented Conversational Response Generation (https://www.aclweb.org/anthology/P19-1371)

Requirements

• python
• pytorch
• torchtext
• argparse
• codecs
• shutil
• re
• progressbar
• collections
• sklearn
• gensim

Configuration

./config.yml is a configuration file, which contains configurations of model, training, and tesing.

1.Preprocessing

Edit the script file build_vocab.sh.

sh scripts/build_vocab.sh

parameter	description
-train_data FILE	Training Set (Source + Target)
-save_data STR	the Prefix of Output File Name
-config FILE	Configuration File

2.Training

sh run.sh

parameter	description
-gpuid INT	Choose Which GPU A Program Uses
-config FILE	Configuration File
-config_with_loaded_model PATH	The config file when loading a warmstart model with its config
-config_from_local_or_loaded_model BOOL	load warmstart model from local path or path mentioned in condig file
-train_data FILE	input of Training
-out_dir FILE	Output model dir
-vocab FILE	Vocabulary (output from Preprocessing)

3.Testing

sh generation.sh

parameter	description
-gpuid INT	Choose Which GPU A Program Uses
-test_data FILE	test file
-test_out FILE	output file
-config FILE	Configuration File
-config_with_loaded_model PATH	The config file when loading a warmstart model with its config
-config_from_local_or_loaded_model BOOL	load warmstart model from local path or path mentioned in condig file
-model FILE	load existing model
-vocab FILE	Vocabulary
-beam_size INT	size for beam search
-decode_max_length INT