In the generation of captions, specifically in the function generate_predictions, you input a dictionary in the blip model. This dictionary has keys of 'image' and 'prompt' where for the prompt value you input the prompts.blip_prompt. There is not a prompt argument in the lavis library for the generate function. When you exclude the prompt key and value from the generate function you get the same caption as output. Am I missing something? Thank you in advance =]
In the generation of captions, specifically in the function generate_predictions, you input a dictionary in the blip model. This dictionary has keys of 'image' and 'prompt' where for the prompt value you input the prompts.blip_prompt. There is not a prompt argument in the lavis library for the generate function. When you exclude the prompt key and value from the generate function you get the same caption as output. Am I missing something? Thank you in advance =]