import torch
from PIL import Image
import open_clip

model, _, preprocess = open_clip.create_model_and_transforms('ViT-B-32', pretrained='laion2b_s34b_b79k')
model.eval()
tokenizer = open_clip.get_tokenizer('ViT-B-32')

image = preprocess(Image.open("source.webp")).unsqueeze(0)
classes = [
    "a bird",
    "a plane",
    "a helicopter",
    "superman"
]
text = tokenizer(classes)

with torch.no_grad(), torch.autocast("cuda"):
    image_features = model.encode_image(image)
    text_features = model.encode_text(text)
    image_features /= image_features.norm(dim=-1, keepdim=True)
    text_features /= text_features.norm(dim=-1, keepdim=True)

    text_probs = (100.0 * image_features @ text_features.T).softmax(dim=-1)[0]

best_guess = torch.argmax(text_probs)

print(classes[best_guess] if text_probs[best_guess] > 0.1 else "unknown")