search

facebookresearch / InferSent

InferSent sentence embeddings
Other
2.28k stars 471 forks source link

How to train InferSent on Urdu dataset #80

Closed MFarhatUllah closed 6 years ago

MFarhatUllah commented 6 years ago

hello, There is any tutorial to train InferSent on any language other then English??? If not then how it is possible?? For Example: If I want train InferSent on Urdu Dataset can I use train_nli.py then how I use it. what chamges are necessory to train it??

MFarhatUllah commented 6 years ago

train_nli.py has 'gpu' code implementation. But i have cpu. how to run it on cpu what changes are needed to run it on 'cpu'.. Please help ....

aconneau commented 6 years ago

Hi, for now InferSent can only be trained on English. However we are soon going to release a corpus/method that will include multiple languages, including Urdu. So stay tuned! About the CPU, unfortunately these LSTM models are too big to be trained on CPU. It would simply be too slow. Once the model is trained however, at inference time, you can generate the embeddings on CPU. Best Alexis

guotong1988 commented 5 years ago

This project do not support training anymore. So do you know a better code for sentence embedding? Thank you!