A clean implementation of "Adversarially Regularized Autoencoders (ICML 2018)" by Zhao, Kim, Zhang, Rush and LeCun https://arxiv.org/abs/1706.04223
To evaluate a model you can download a pretrained kenlm model (the model trained on the same train.txt file):
python train.py --data data_snli --no_earlystopping --gpu 0 --kenlm_model knlm_snli.arpa
- After 1 epoch:
- the people are looking at food .
- the little boy is going to catch the grass .
- the old man is not being .
- the man and woman are having in a kitchen .
- four girls walk along a street while others watch
- After 2 epochs:
- two women are in a crowd .
- couple sitting with their women working .
- a man can be walking through the grass in the mountains .
- a man is trying to buy down a wall
- women in a large lab room
- After 5 epochs:
- a group of adults with a clean up .
- a bike is decorated in a mountain station .
- people are playing rugby .
- a man is in his hand of a t-shirt .
- a man tries to prepare for the lake .
- After 10 epochs:
- a basketball player dancing on the beach
- a man at a tall go getting .
- a old woman his scooter before racing .
- the female is wearing red bicycle for a snowmobile .
- a little girl in a red scarf is bed and sleeping on his room .
| option | description |
|---|---|
| --tensorboard | draw graphs. need tensorboardx to work |
| --kenlm_model | path to reference kenlm model for computing forward ppl |
| --gpu | -1 - don't use gpu, > -1 - use |
| --compressing_rate | -S param for kenlm cmd line util |
python generate.py --greedy