Can not reproduce the results of "Finetuning with NOLA (float)" #1
Description
Hey @klnavaneet, thank you for your nice work! I follow the instruction of "Finetuning with NOLA (float)" and run the codes, and finally the results are
- BLEU: 0.6977
- NIST: 8.7974
- METEOR: 0.4656
- ROUGE_L: 0.7148
- CIDEr: 2.4907,
which are a bit lower than those mentioned in your paper.
And I'd greatly appreciate it if you could share the latest version of codes. I'm wondering if I made any mistakes while editing the codes.(There seems to be some bugs in the current version, and I attempted to correct them. Maybe I messed up here.😂)