Pmpico/blip-api

1from PIL import Image2import requests3import gradio as gr4 5from transformers import BlipProcessor, BlipForConditionalGeneration6 7model_id = "Salesforce/blip-image-captioning-base"8 9model = BlipForConditionalGeneration.from_pretrained(model_id)10processor = BlipProcessor.from_pretrained(model_id)11 12def launch(input):13    image = Image.open(requests.get(input, stream=True).raw).convert('RGB')14    inputs = processor(image, return_tensors="pt")15    out = model.generate(**inputs)16    return processor.decode(out[0], skip_special_tokens=True)17 18iface = gr.Interface(launch, inputs="text", outputs="text")19iface.launch()