




## Q1. how to run the code? 
    see local_train.sh which runs ./train.py  
    ## because we cannot upload the data, thus the code cannot be run directly;


## Q2: how is the the GR-model like ?
    the model structrue see  ./models/model_exp2.py
        when you want to use multi-task GR model , set 
            hs_baseline = True,    ## direct output hs embedding
            single_encoder = True,     ## use only one context encoder
        
        
        if you want to train Seperate Context encoder GR model , set 
            hs_baseline = False,        ## use (hs-hc) in the gloss matching 
            single_encoder = False,     ## use seperate context encoder 

## Q3: how is the data like ?
    we gives examples of the data in the folder ./examples_train_data
    the id_format file "example_roberta_ids.txt" is the input of the model, which will be processed in the code ./models/dataset_v2.py,
    which will process the data for the model to train/eval 