-
## 一言でいうと
翻訳における、逆翻訳の有効性について検証した研究。逆翻訳では、学習する翻訳が日=>英の場合、英のみのコーパスを英=>日に戻す(逆翻訳)ことでデータの水増しを行う。逆翻訳時に、greedyでなくノイズを加えた方が良好な結果。これにより商用の翻訳サービスを上回る結果
### 論文リンク
https://arxiv.org/abs/1808.09381
###…
-
## 🐛 Bug
Using the **--print-alignment** argument makes the generation up to 3x times slower (Depends on the batch size). For example, generating translation for my test set took 47.9s (134.54 sent…
-
### Problem Description
File "/opt/conda/envs/py_3.9/lib/python3.9/site-packages/fairseq/optim/fp16_optimizer.py", line 393, in backward
loss.backward()
File "/opt/conda/envs/py_3.9/lib/pyt…
-
Hi,
when I run"evaluate_vqa_beam.sh test'' the error appear:
```
Show Dotfiles Show Owner[/](https://ood.hpc.nyu.edu/pun/sys/files/fs/)Mode
/[scratch](https://ood.hpc.nyu.edu/pun/sys/files/fs/sc…
-
Hi, I looked all over the git, but I couldn't find the code for the normalization?
Can you help me out here?
Thanks.
-
请问您的transformer-base的序列长度是用的默认的1024吗？因为pdc的测试集中，中文的148个篇章中有35个是超过了1024（按子词计算），那么如果采用1024，这35个篇章该怎么处理呢？谢谢
-
## 🐛 Bug
Have traceback with error `Error(s) in loading state_dict for Wav2VecCtc:`
On commit 68157726 speech recognition example worked. on last changes have traceback.
I use infer.py from s…
-
import fairseq
import torch
import torchaudio
import matplotlib.pyplot as plt
from scipy.spatial.distance import cdist
wav2vec2_checkpoint_path = "xlsr_53_56k.pt"
checkpoint = torch.load(wav2vec…
-
## ❓ Questions and Help
#### What is your question?
I came across this assertion: https://github.com/facebookresearch/fairseq/blob/3f6ba43f07a6e9e2acf957fc24e57251a7a3f55c/fairseq/modules/multihea…
-
## 🐛 Bug
### To Reproduce
Steps to reproduce the behavior (**always include the command you ran**):
1. Run cmd 'pip install --editable ./'
2. See error
```
/usr/bin/clang -Wno-unused…