Script for running Florence2 based models

run_florence2_simple.py

import requests
import torch
from PIL import Image
from transformers import AutoModelForCausalLM, AutoProcessor

device = "cuda:0" if torch.cuda.is_available() else "cpu"
torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32

model_id = r"microsoft/Florence-2-base" # can also use path/to/huggingface/hub/florence2-repo-name/snapshots/hash or something like that
model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch_dtype, trust_remote_code=True).to(device)
processor = AutoProcessor.from_pretrained(model_id, trust_remote_code=True)

prompt = "<CAPTION>"
# prompt = "<OCR>" # theres a lot more, look it up on HF

image_path = r"path/to/image.png"
image = Image.open(image_path).convert("RGB")

inputs = processor(text=prompt, images=image, return_tensors="pt").to(device, torch_dtype)

generated_ids = model.generate(
    input_ids=inputs["input_ids"],
    pixel_values=inputs["pixel_values"],
    max_new_tokens=4096,
    num_beams=3,
    do_sample=False,
)
generated_text = processor.batch_decode(generated_ids, skip_special_tokens=False)[0]

# task = prompt here because too lazy to give it unique thingy
parsed_answer = processor.post_process_generation(generated_text, task=prompt, image_size=(image.width, image.height))

print(parsed_answer)