Instructor patched SDK worked good as well, but this just cleans up the code and is simple. Based on my own implementation with instructor, it obeys the Pydantic basemodel structures close to a 100% of the completions, compared to the few-shot propmting setup i had before that, that had about 10% success and LOADS of failures 🙂 This way just works !
But given that OpenAI, Claude can do this already, why do we need ollama for it? And here, isn’t Llama doing the heavy lifting? Not clear how ollama is helping.
![How to Get Structured Outputs with Ollama [Pydantic Models]](http://i.ytimg.com/vi/KCJwvMFLjgA/mqdefault.jpg)
Awesome video. I’m from America and I visited Wales and England earlier this year for the first time. Wales was so beautiful.
Ive been needing this feature, thanks for the info!
Great adavancment for open source efforts
I’d appreciate if you could share the codes.
Sure thing, I've added a link with them to the description
@@IanWootten Diolch yn fawr
Instructor patched SDK worked good as well, but this just cleans up the code and is simple.
Based on my own implementation with instructor, it obeys the Pydantic basemodel structures close to a 100% of the completions, compared to the few-shot propmting setup i had before that, that had about 10% success and LOADS of failures 🙂 This way just works !
👋👋👋
But given that OpenAI, Claude can do this already, why do we need ollama for it? And here, isn’t Llama doing the heavy lifting? Not clear how ollama is helping.
Using a model that's local, open and free are three good reasons.