Yang An

@qyc-98 @RishabhMaheshwary @ilovecv Hi, we have found the bug and fixed it! Now the latest codebase supports open-ended (unconstrained) VQA finetuning and evaluation. Please pull the latest code and refer...

@leng-yue We have tested open-ended VQA fine-tuning on OFA-base (without using EMA). It achieves 76.4 score on our VQA validation set. This performance can still be improved by using EMA...

Meanwhile, the file `train+val2014_qla_mrcnn.json` to finetune on both the train and valid set is also missing.

@maxinge8698 No I haven't found these files. Now I have managed to implement my own model, which is named OFA and have achieved a better accuracy than VinVL (test-std 80.02)...

@Lizw14 Hi, may I ask could you please provide mask-rcnn labels of the test-dev and test-std set? Thank you very much!

Excuse me, may I ask whether you have these files `train+val2014_qla_mrcnn.json`, `test2015_qla_mrcnn.json` and `test-dev2015_qla_mrcnn.json`? I found these files are missing, making it difficult for inference and official evaluation.

In this closed issue (#13), I noticed the author has mentioned the way to generate the mask-rcnn-based object labels. I tried to reproduce the labels on the VQA training images....

Hi, while finetuning on flickr30k, did you encounter this error? `RuntimeError: expand(torch.cuda.FloatTensor{[64, 1, 4, 100, 2048]}, size=[64, 4, 4, 2048]): the number of sizes provided (4) must be greater or...

Hi, could you please release the features from a small sampled set of Conceptual Caption images? It will be much helpful for us to check the correctness of the computed...

@zhouwei5113 @Zasder3 Hi, maybe you can refer to this repo! https://github.com/OFA-Sys/Chinese-CLIP