Sanghwan Jang comments

Repositories
Issues
Comments

Results 2 comments of


                                            Sanghwan Jang

BertLMHeadModel (w/ relative position embedding) does not work correctly when use_cache = True

Thank you for your quick reply, of course! With absolute position embedding (`position_embedding_type = 'absolute'`), the minimal code sample outputs: ``` tensor([[[[0.2540, 0.2539, 0.2472, 0.2449]], [[0.2519, 0.2480, 0.2483, 0.2518]], [[0.2473,...

BertLMHeadModel (w/ relative position embedding) does not work correctly when use_cache = True

@ydshieh Ah... I forgot to tag you. I think I can fix this problem. If you agree that the above should be handled and you're not working on it, I'll...