VQA algorithm data loader have no truncation for answer

OFA-Sys / OFA

Official repository of OFA (ICML 2022). Paper: OFA: Unifying Architectures, Tasks, and Modalities Through a Simple Sequence-to-Sequence Learning Framework

Apache License 2.0

2.39k stars 248 forks source link

VQA algorithm data loader have no truncation for answer #365

Open TheOnlylight opened 1 year ago

TheOnlylight commented 1 year ago

I am trying to use vqa on a task to extract information from text-riched image. The thing is that in the cases that answer is too long the model would report a mismatch in tensor size. I checked the data design and didn't find a truncation setting for the answer data. How can I solve this.

The size of tensor a (1161) must match the size of tensor b (1024)

yangapku commented 1 year ago

Hi, can you provide more details about the samples which caused the errors?