What's the best performance of this code?

[gorgeousyouth]

excuse me , i get the model by using the code , but i can not get the score in the paper,can you tell me the reason

[weicheng51]

Bleu_1: 0.639145
Bleu_2: 0.424485
Bleu_3: 0.283304
Bleu_4: 0.192496
METEOR: 0.203724
ROUGE_L: 0.503118
CIDEr: 0.584677

[weicheng51]

the result of my training is far from the one published in the paper

[shiqi1994]

the result of my training is far from the one published in the paper

I got the same results which indeed far away from the results in the paper. And I do not know why:(

[takagi97]

I imitated this code and wrote a Pytorch-based version, and used Resnet101 to extract features, only to get a score close to that in the paper.
ratio: 1.01423555697
Bleu_1: 0.708868328298
Bleu_2: 0.50652071489
Bleu_3: 0.359143843477
Bleu_4: 0.254603597177
METEOR: 0.231556508619