{"type":"video","version":"1.0","html":"<iframe src=\"https://www.loom.com/embed/6aa43afa75144decbc8dec564e24bdbc\" frameborder=\"0\" width=\"1110\" height=\"832\" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>","height":832,"width":1110,"provider_name":"Loom","provider_url":"https://www.loom.com","thumbnail_height":832,"thumbnail_width":1110,"thumbnail_url":"https://cdn.loom.com/sessions/thumbnails/6aa43afa75144decbc8dec564e24bdbc-eea8a914b47fe920.gif","duration":331.319,"title":"Sutro Demo: Synthetic Data Generation","description":"In this video, I demonstrate how to use Sutro, an accelerated batch inference service that allows you to run thousands or even millions of LLM inference calls quickly and inexpensively. I walk through generating synthetic data, specifically artificial code review comments, which can be beneficial for training reward models or distilling capabilities into smaller LLMs. We import the Sutro Python library, load the starcoder dataset, and create prompts to generate comments efficiently. I encourage you to explore Sutro's capabilities, whether through the Python library or the web UI, to refine your models and parameters. Please take a look at the demo and consider how this tool can enhance your research workflow."}