-
Type:
Task
-
Resolution: Unresolved
-
Priority:
Unknown
-
None
-
Affects Version/s: None
-
Component/s: ai-ml-testing, AI/ML, LangChain
Context
Describe the background behind the problem.
In addition to a "temperature" setting, it has come to my attention that the ChatOpenAI class (as well as the Azure one, also derived from `BaseChatOpenAI`) has a random seed argument!
It is my hope that we can use this to reduce the flakiness of our tests.
Definition of done
What must be done to consider the task complete?
Investigate whether adding `seed=0` wherever we create an LLM in our tests leads to consistent output. Update throughout tests used in ai-ml testing pipelines.
Primary focus:
- LangChain{}
Pitfalls
What should the implementer watch out for? What are the risks?