Update README.md

fix tokenizer bug

fix readme

replace with new dataset :)

Add link to blog

support evaluating with several answers

improve evaluation

small log fix and workers=0 to avoid bug

Create validation test for VQAInstruct

slurm job for first round of ablations

dont evaluate on the whole evaluation data

add multiple evaluation datasets

add wandb

added LORA training

added VQA dataset