Inference cost for an AI image feature doubled without usage growth. Diagnose it.

Focus on identifying potential factors that could lead to a doubling of inference costs without a corresponding increase in usage. Consider aspects such as changes in the underlying model architecture, increased computational resource requirements, or inefficiencies in the deployment environment. Explore whether there have been any recent updates or changes in the data pipeline, model version, or infrastructure that could have impacted performance. Additionally, think about external factors like increased cloud service costs or changes in user behavior that might not directly correlate with usage growth but could affect costs. Be prepared to prioritize your findings and suggest actionable steps for further investigation or mitigation.