pytorch 1.0.0 implementation of semantically conditioned VAE
Paper: https://arxiv.org/abs/1812.08879
See run-example.sh or use the following commands for quick start
An pre-trained model on all domains is stored in checkpoints/example/cvae_domain-all/
The training log is at log/example, generation result is at res/example/ and bleu result is at bleu/example
model_path='scvae'
log=./scvae.log
res=./scvae.res
bleu=./scvae.bleu
TRAIN
python3 train.py --model_path=$model_path > $log
TEST
python3 test.py --model_path=$model_path > $res
BLEU
python3 score_on_res.py $res 1 > $bleu