Open hackerchenzhuo opened 3 years ago
[{"test-dev": {"yes/no": 87.19, "number": 52.96, "other": 60.83, "overall": 70.79}}, {"test-standard": {"yes/no": 86.82, "number": 53.14, "other": 61.05, "overall": 70.87}}]
this is my result, which is different to 72.42 / 72.54 as the paper shows.
72.42 / 72.54
PS: I modify the code with huggingface Version with:
self.lxmert_tokenizer = LxmertTokenizer.from_pretrained("unc-nlp/lxmert-base-uncased") self.lxmert_vqa = LxmertForQuestionAnswering.from_pretrained("unc-nlp/lxmert-vqa-uncased")
Is it for this reason?
[{"test-dev": {"yes/no": 87.19, "number": 52.96, "other": 60.83, "overall": 70.79}}, {"test-standard": {"yes/no": 86.82, "number": 53.14, "other": 61.05, "overall": 70.87}}]
this is my result, which is different to
72.42 / 72.54
as the paper shows.PS: I modify the code with huggingface Version with:
Is it for this reason?