ViChartVQA A2 - Transformer Decoder
This repository stores the A2 model checkpoint for Vietnamese Chart Visual Question Answering.
Model
A2 uses a custom architecture with an image encoder, text encoder, co-attention fusion, and a Transformer answer decoder.
File
best_model_a2.pth
Note
This repository only stores the PyTorch checkpoint.
The custom model architecture code is required to load this checkpoint.