Error on Evaluating GPT2LMHeadModel after training from scratch

I was training the GPT2LMHeadModel From Scratch with my own dataset. i want to check the accuracy and loss. however when I run trainer. evaluate() it returns the following error. i google for several hours but I couldn’t find a soln.

The Compute Metrix:

def compute_metrics(eval_preds):
    metric = load_metric("glue", "mrpc")
    logits, labels = eval_preds
    predictions = np.argmax(logits, axis=-1)
    return metric.compute(predictions=predictions, references=labels)

The Error

# [TypeError: only size-1 arrays can be converted to Python scalars]