I follow the example provided here
The paper https://arxiv.org/pdf/2212.09662v2.pdf mentions that we can also extract the code form the chart images. But I am not what parameters we need to pass.
I follow the example provided here
The paper https://arxiv.org/pdf/2212.09662v2.pdf mentions that we can also extract the code form the chart images. But I am not what parameters we need to pass.
Could you be specific about what want to do with the model? Is it just to extract the code from chart images?
Prediction of code is part of the models pre-training objective. The models available on huggingface platform are fine-tuned for specific tasks on certain datasets. Hence these models may not be very good at extracting code. You may need to fine-tune it for your specific purpose if you were to have a decent performance.
Thanks, yes, I misunderstood the matcha paper. THey are using it as a pertaining task.