Usage
import requests
from PIL import Image
import torch
from transformers import AutoProcessor, LlavaOnevisionForConditionalGeneration
model_id = "NicoZenith/onevision-7b-all-vqa-conv"
model = LlavaOnevisionForConditionalGeneration.from_pretrained(
model_id,
torch_dtype=torch.float16,
low_cpu_mem_usage=True,
).to(0)
processor = AutoProcessor.from_pretrained(model_id)
conversation = [
{
"role": "user",
"content": [
{"type": "text", "text": "What can you say about this X-ray?"},
{"type": "image"},
],
},
]
prompt = processor.apply_chat_template(conversation, add_generation_prompt=True)
image_file = "https://prod-images-static.radiopaedia.org/images/29923576/fed73420497c8622734f21ce20fc91_gallery.jpeg"
raw_image = Image.open(requests.get(image_file, stream=True).raw)
inputs = processor(images=raw_image, text=prompt, return_tensors='pt').to(0, torch.float16)
output = model.generate(**inputs, max_new_tokens=200, do_sample=False)
response_text = processor.decode(output[0][2:], skip_special_tokens=True)
response_text = response_text.split("assistant\n")[-1]
print(response_text)
