import spaces
from transformers import pipeline as tpipeline
# from optimum.pipelines import pipeline as opipeline

#@spaces.GPU(duration=60)
def classify(tweet, event_model, hftoken, threshold):
    results = {"text": None,  "event": None, "score": None}
    
    # event type prediction with transformers pipeline
    event_predictor = tpipeline(task="text-classification", model=event_model, 
                               batch_size=512, token=hftoken, device="cpu")
    tokenizer_kwargs = {'padding': True, 'truncation': True, 'max_length': 512}
    prediction = event_predictor(tweet, **tokenizer_kwargs)[0]


    # with onnx pipeline 
    # onnx_classifier = opipeline("text-classification", model=event_model, accelerator="ort", 
    #                             batch_size=512, token=hftoken, device="cpu")
    # prediction = onnx_classifier(tweet)[0]
    
    
    results["text"] = tweet
    
    if prediction["label"] != "none" and round(prediction["score"], 2) <= threshold:
        results["event"] = "none"
        results["score"] = prediction["score"]
    else:
        results["event"] = prediction["label"]
        results["score"] = prediction["score"]

    return results