• Suraj Patil's avatar
    [examples/seq2seq] support label smoothing (#9844) · 1cd16512
    Suraj Patil authored
    * add prepare_decoder_input_ids_from_labels in s2s models
    
    * support lbl smoothing and enc/emb freezing
    
    * fix freezing
    
    * use pad_token_id from config
    
    * remove embed freezing and add warning
    
    * prepare decoder_input_ids inside DataCollatorForSeq2Seq
    1cd16512
modeling_mbart.py 80.8 KB