BLEU Score Computation

[sleighsoft]

I have two questions:

1. How do you compute the reported BLEU scores? I see the imports of BLEU in several models but BLEU is never added to the self.metrics collection of any of them.

2. When computing BLEU, which sentences are considered to be references?

[sleighsoft]

@Yaoming95 I would really love to replicate your results.

[wyxcc]

@Yaoming95 I would really love to replicate your results.

Have you solved the problem? :)

[sleighsoft]

No. To be honest, I stopped pursuing BLEU as a solid score.

[khorshidisamira]

Hello, How did you add BLEU metric to your metrics?

[shizhediao]

Same question here. @Yaoming95

[xuefeng199507]

@Yaoming95 I would really love to replicate your results.

Have you solved the problem? :)

Hello, I encountered the same problem as you, please tell me how you solved it?Thanks

[tangshu-yu]

from utils.metrics.Bleu import Bleu

if name == 'main': bleu_metric = Bleu(test_text=r'/docs/instances/seq.txt', real_text=r'data/image_coco.txt', gram=2) score = bleu_metric.get_score(is_fast=False, ignore=False) print(f"BLEU Score: {score}")