I get different answer using Hosted inference API and pipeline object

Hi i trained convbert base to do question answering. When I use it with pipeline object it’s giving me different answers than API aveliable on your website

model = TFAutoModelForQuestionAnswering.from_pretrained('Gozdi/convbert-base-adversarial-qa-exp-1-epoch-2')

qa_pipeline = pipeline(task='question-answering', model=model, tokenizer=tokenizer)

qa_pipeline(question="Where do I live?", context='My name is Wolfgang and I live in Berlin')

{'score': 0.7279412746429443, 'start': 34, 'end': 40, 'answer': 'Berlin'}