search

microsoft / GLUECoS

A benchmark for code-switched NLP, ACL 2020
https://microsoft.github.io/GLUECoS
MIT License
73 stars 58 forks source link

Trained Models #68

Closed agarwalchaitanya closed 2 years ago

agarwalchaitanya commented 2 years ago

Are the trained models on each task available publicly? Is it possible for the authors to share them? Specifically, I'm looking for code-mixed NMT and LID.

TIA!

Genius1237 commented 2 years ago

Nope. You can however look at the huggingface model hub link, and you'll find that some users have uploaded their own custom models trained on different code-mixed datasets.

agarwalchaitanya commented 2 years ago

Nope. You can however look at the huggingface model hub link, and you'll find that some users have uploaded their own custom models trained on different code-mixed datasets.

I couldn't find anything substantial on hf. I'm thinking of retraining models anyway but can't find a named script for LID. Can you direct me to it? And I guess for NMT, you'd be using fairseq?

Genius1237 commented 2 years ago

LID uses the BertToken.py script. You can see how to launch it and pass arguments to it in the train_token.sh file. For MT, we use a huggingface based script and this is there at run_seq2seq.py. This gets launched from train_mt.sh, so you can look at that for how to pass the arguments. All these above mentioned scripts are in the Code directory.