oyoy
oyoy
请问这个问题你是怎么解决的呢？直接将它模型定义部分的代码全部注释替换成AutoModel.pretrain吗
By the way, the vqa_val.tsv which download from #68 could run evaluate_vqa_beam.sh on multiple GPU, but my fine only could run on single GPU.
Hey, thanks for your issue. It's easy to work out. You just need to add a '/ ' after the model's path.
We are considering updating the code for the aok-vqa MC part. You also can try to change the prompt code of Aok-vqa DA part according the [paper](https://arxiv.org/pdf/2303.01903)'s Table 11.
You can easily change Stage 2’s language model to your own local model by modifying the inference code, but the required environment could be change.
Yes，It's correct format for the example files. The error seems like you are using a okvqa's file while running aokvqa's task. Please make sure that the candidate files is aok's...