This folder contains the code for the paper 《Style Transformer: Unpaired Text Style Transfer without Disentangled Latent Representation》
- Requirements
- pytorch >= 0.4.0
- torchtext >= 0.4.0
- nltk
- fasttext == 0.8.3
- kenlm
- nltk
- Run the following command to download (or double-check) nltk tokenizer dependency:
python -c "import nltk; nltk.download('punkt')" - LM for evaluator
- download from ppl_yelp.bin
- put
ppl_yelp.binaryin the folderevaluator
The hyperparameters for the Style Transformer can be found in ''main.py''.
The most of them are listed below:
data_path : the path of the datasets
log_dir : where to save the logging info
save_path = where to save the checkpoing
discriminator_method : the type of discriminator ('Multi' or 'Cond')
min_freq : the minimun frequency for building vocabulary
max_length : the maximun sentence length
embed_size : the dimention of the token embedding
d_model : the dimention of Transformer d_model parameter
h : the number of Transformer attention head
num_layers : the number of Transformer layer
batch_size : the training batch size
lr_F : the learning rate for the Style Transformer
lr_D : the learning rate for the discriminator
L2 : the L2 norm regularization factor
iter_D : the number of the discriminator update step pre training interation
iter_F : the number of the Style Transformer update step pre training interation
dropout : the dropout factor for the whole model
log_steps : the number of steps to log model info
eval_steps : the number of steps to evaluate model info
slf_factor : the weight for self reconstruction loss
cyc_factor : the weight for cycle reconstruction loss
adv_factor : the weight for style controlling loss
You can adjust them in the Config class from the ''main.py''.
If you want to run the model, use the command:
python main.pyUpdate: You can find the outputs of our model in the "outputs" folder. here is the ckpt for imdb and yelp dataset: ckpt.rar
******************** neg sample ********************
[gold] there chips are ok , but their salsa is really bland .
[raw ] there chips are ok , but their salsa is really bland .
[rev ] pub chips are ok but their salsa is really good !
[ref ] these chips are okay but their salsa is really tasty .
[gold] blue cheese dressing was n't the best by any means .
[raw ] blue cheese dressing was n't the best by any means .
[rev ] blue cheese dressing was the best means !
[ref ] blue cheese dressing was above average .
******************** pos sample ********************
[gold] amazing food , great service .
[raw ] amazing food , great service .
[rev ] terrible hostess , horrible kfc .
[ref ] lousy food , slow service .
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent