Evaluation Code about clip_prefix_caption HOT 5 CLOSED

We use Karpathy et al. split for COCO, and the code of Oscar repo
For conceptual captions we use the validation set as the test set wasn't released and require evaluation on Google Cloud, but we didn't use the validation set for validation purposes at all.

Hope it helps
@amirhertz

from clip_prefix_caption.

@rmokady Thank you for answering!

Where can I find details about the karpathy split, how each data split is created? I saw you uploaded the train_captions but not val/test captions

from clip_prefix_caption.

We just use the same split as used in Oscar repo

You can run their inference script which will produce a json with test captions

from clip_prefix_caption.

We just use the same split as used in Oscar repo

You can run their inference script which will produce a json with test captions

Could you please share your val/test captions? I'm having some difficulty running Oscar.

from clip_prefix_caption.

我们只使用与 Oscar 存储库中使用的相同的分割
您可以运行他们的推理脚本，该脚本将生成带有测试标题的 json

您能分享您的验证/测试标题吗？我在竞选奥斯卡时遇到了一些困难。

请问您解决了吗这个我呢提，我也遇到了这个问题

from clip_prefix_caption.