Potential bug in predict.py

[chawannut157]

In predict.py, it's still using use_cuda=args.cuda, which is no longer used in the eval_model function.

eval_model(model, test_data,metric_meta=metric_meta,**use_cuda=args.cuda,** with_label=args.with_label)

In inference.py

def eval_model(model, data, metric_meta, device, with_label=True, label_mapper=None, task_type=TaskType.Classification):

[namisan]

thanks @chawannut157 , fixed the bug.