Comments (5)
Hello,
Can you have a look at #13 or #14 ? If this doesn't solve your problem, please tell me.
from unsupervisedmt.
Thank you very much. It‘s finally work!And on newstest2014 en-fr, it give above 23.33 BLEU after 25 epochs. But there is another question. When I change my translation task to Mongolian-Chinese,I modified the /NMT/get_data.sh and got some data (/data/mono/all.mn.tok.6000.pth, /data/mono/all.zh.tok.6000.pth, /data/para/dev/val1w.XX.6000.pth, /data/para/dev/test1w.XX.6000.pth, /data/mono/all.mn-zh.6000.vec).
Then I run the command as follows:
python main.py --exp_name mn-zh --transformer True --n_enc_layers 3 --n_dec_layers 3 --share_enc 2 --share_dec 2 --max_len=100 --share_lang_emb True --share_output_emb True --langs 'mn,zh' --n_mono -1 --mono_dataset 'mn:./data/mono/all.mn.tok.6000.pth,,;zh:./data/mono/all.zh.tok.6000.pth,,' --para_dataset 'mn-zh:,./data/para/dev/val1w.XX.6000.pth,./data/para/dev/test1w.XX.6000.pth' --mono_directions 'mn,zh' --word_shuffle 3 --word_dropout 0.1 --word_blank 0.2 --pivo_directions 'zh-mn-zh,mn-zh-mn' --pretrained_emb './data/mono/all.mn-zh.6000.vec' --pretrained_out True --lambda_xe_mono '0:1,100000:0.1,300000:0' --lambda_xe_otfd 1 --otf_num_processes 8 --otf_sync_params_every 1000 --enc_optimizer adam,lr=0.0001 --epoch_size 20 --stopping_criterion bleu_mn_zh_valid,1
But the Bleu is 0, I want to know if NMT can be used for Mongolian-Chinese translation tasks. What should I do?
from unsupervisedmt.
You are only using 6000 BPE codes is that expected? Usually I use 60000. I tried 8000 once and it gave me pretty bad results too. Also, did you evaluate the word translation accuracy of your cross-lingual BPE embeddings? If you had a mn-zh dictionary you could evaluate the word translation accuracy using https://github.com/facebookresearch/MUSE or something.
Also, how big are your monolingual corpora? It looks like the training epoch finished in 8 seconds. You also set --epoch_size 20
, why? Usually epoch size is around 100000 or something.
from unsupervisedmt.
The number of monolingual sentences for each language is 1230000,and when I used 60000 BPE codes, CUDA error: out of memory is come again,so I changed it to 6000.
I will try to evaluate the word translation accuracy using https://github.com/facebookresearch/MUSE . Thank you!
from unsupervisedmt.
Hi,I run the unsupervised NMT codes and PBSMT codes,but the BLEU are both 0. Is there too little monolingual corpus?
NMT codes:
python main.py --exp_name mn-zh --transformer True --n_enc_layers 2 --n_dec_layers 2 --share_enc 1 --share_dec 1 --max_len=50 --share_lang_emb True --share_encdec_emb True --share_decpro_emb True --share_output_emb True --langs 'mn,zh' --n_mono -1 --mono_dataset 'mn:./data/mono/all.mn.tok.40000.pth,,;zh:./data/mono/all.zh.tok.40000.pth,,' --para_dataset 'mn-zh:,./data/para/dev/val3k.XX.40000.pth,./data/para/dev/test3k.XX.40000.pth' --mono_directions 'mn,zh' --word_shuffle 3 --word_dropout 0.1 --word_blank 0.2 --pivo_directions 'zh-mn-zh,mn-zh-mn' --pretrained_emb './data/mono/all.mn-zh.40000.vec' --pretrained_out True --lambda_xe_mono '0:1,100000:0.1,300000:0' --lambda_xe_otfd 1 --otf_num_processes 8 --otf_sync_params_every 1000 --enc_optimizer adam,lr=0.0001 --epoch_size 100000 --stopping_criterion bleu_mn_zh_valid,10 --batch_size 32
from unsupervisedmt.
Related Issues (20)
- why MemoryError
- Why codes file is empty.? HOT 4
- for different language, where to make change?
- How to train NMT + PBSMT ?
- UnboundLocalError: local variable 'n_words' referenced before assignment
- About number of shared layers
- RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.cuda.FloatTensor [14, 32, 1536]], which is output 0 of AddBackward0, is at version 2; expected version 1 instead. Hint: enable anomaly detection to find the operation that failed to compute its gradient, with torch.autograd.set_detect_anomaly(True). HOT 1
- How to run PBSMT +NMT ?
- transformer multihead attention scaling layer error
- Setting the random seed does not result in same outputs across runs
- I have trouble when run get_data_enfr.sh
- How can I modify the code to train may own dataset on specific language?
- Low utilization rate of cuda HOT 1
- How to train the vector of phrases
- Low BLEU on PBSMT HOT 3
- bpe_end issue
- Getting raise EOFError() while executing Linux Command through Netmiko
- How i can run MUSE alignment in .sh
- How to train the model without para_dataset
- Error in runny bash command. HOT 2
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from unsupervisedmt.