app.py

import torch
from transformers import CLIPProcessor, CLIPTokenizer, CLIPModel

# Load pre-trained CLIP model and processor
model = CLIPModel.from_pretrained("clip-vit-base-patch16-224", torch.device("cpu"))
processor = CLIPProcessor.from_pretrained("clip-vit-base-patch16-224")

# Define a function to generate an image from a text prompt
def generate_image(prompt, height=256, width=256):
    # Preprocess the input text
    inputs = processor(text=prompt, return_tensors="pt")

    # Forward pass
    outputs = model(**inputs)

    # Get the image embeddings
    image_embeddings = outputs.image_embeddings

    # Generate an image from the embeddings
    image = torch.randn(1, 3, height, width)
    image = image.to(torch.device("cpu"))
    for i in range(10):  # 10 iterations
        image = model.get_input_embeddings()(image) + image_embeddings
        image = torch.clamp(image, -1, 1)

    # Convert the image to PIL format
    from PIL import Image
    image = image.detach().numpy()
    image = image.transpose(0, 2, 3, 1)
    image = (image + 1) / 2
    image = Image.fromarray((image * 255).astype("uint8"))

    return image

# Test the function
prompt = "A picture of a cat"
image = generate_image(prompt)
image.save("cat.png")