app.py

import gradio as gr
import spaces

import torch
torch.jit.script = lambda f: f
# torch.autocast = lambda device_type, dtype: torch.autocast(device_type, torch.float)

from t2v_metrics import VQAScore, list_all_vqascore_models

print(list_all_vqascore_models())

# Initialize the model only once
# if torch.cuda.is_available(): 
model_pipe = VQAScore(model="clip-flant5-xl", device="cpu")  # our recommended scoring model
print("Model initialized!")

@spaces.GPU
def generate(model_name, image, text):
    # print("Model_name:", model_name)
    print("Image:", image)
    print("Text:", text)
    # model_pipe = VQAScore(model="clip-flant5-xl")  # our recommended scoring model
    # print("Model initialized, now moving to cuda")
    # model_pipe.to("cuda")
    print("Generating!")
    # with torch.autocast(device_type='cuda'):
    # with torch.autocast(device_type='cuda', dtype=torch.float):
    #     result = model_pipe(images=[image], texts=[text])
    #     return result
    return 10

iface = gr.Interface(
    fn=generate,  # function to call
    inputs=[gr.Dropdown(["clip-flant5-xl", "clip-flant5-xxl"], label="Model Name"), gr.Image(type="filepath"), gr.Textbox(label="Prompt")],  # define the types of inputs
    # inputs=[gr.Image(type="filepath"), gr.Textbox(label="Prompt")],  # define the types of inputs
    outputs="number",  # define the type of output
    title="VQAScore",  # title of the app
    description="This model evaluates the similarity between an image and a text prompt."
).launch()