Skip to content

Commit

Permalink
Update README.md
Browse files Browse the repository at this point in the history
  • Loading branch information
jadore801120 committed Aug 27, 2018
1 parent 7e14834 commit 24e3d8d
Showing 1 changed file with 9 additions and 18 deletions.
27 changes: 9 additions & 18 deletions README.md
Expand Up @@ -78,27 +78,18 @@ python translate.py -model trained.chkpt -vocab data/multi30k.atok.low.pt -src d
## Training

<p align="center">
<img src="https://imgur.com/A6byNQ8.png" width="400">
<img src="https://imgur.com/5z4f4tw.png" width="400">
<img src="https://imgur.com/rKeP1bb.png" width="400">
<img src="https://imgur.com/9je3X6U.png" width="400">
</p>

- Parameter settings:
- batch_size=64
- d_inner_hid=1024
- d_k=64
- d_v=64
- d_model=512
- d_word_vec=512
- dropout=0.1
- embs_share_weight=False
- n_head=8
- n_layers=6
- n_warmup_steps=4000
- proj_share_weight=True

- Elapse per epoch (on NVIDIA Titan X):
- Training set: 1.38 min
- Validation set: 0.016 min
- default parameter and optimizer settings
- label smoothing
- target embedding / pre-softmax linear layer weight sharing.

- Elapse per epoch (on NVIDIA Titan X):
- Training set: 0.888 minutes
- Validation set: 0.011 minutes

## Testing
- coming soon.
Expand Down

0 comments on commit 24e3d8d

Please sign in to comment.