handler.py

from transformers import CLIPModel, CLIPProcessor
import torch
import numpy as np

device = "cuda" if torch.cuda.is_available() else "cpu"

model = CLIPModel.from_pretrained(".").to(device)
processor = CLIPProcessor.from_pretrained(".")

def inference(inputs):
    text = inputs.get("text", None)

    if text:
        processed = processor(text=[text], return_tensors="pt", truncation=True).to(device)
        with torch.no_grad():
            emb = model.get_text_features(**processed)
        emb = torch.nn.functional.normalize(emb, dim=-1)
        return emb.cpu().numpy().tolist()

    return {"error": "No valid input provided"}