Text-conditional image generation model based on OpenAI's unCLIP