mlflow - connecting endpoints to tracking server #277
Conversation
tayyib-saddique
changed the title
|
Where is the tracking server storing experiments? Is the intention to deploy this so we can all see each other's experiments, and will this setup work that way? |
Tracking server stores experiments locally - have discussed potential possibility to deploy in AWS so experiments can be shared in line with broader work related to DeepEval and Ragas |
tayyib-saddique
force-pushed
the
experiment-mlflow-prompt-testing
branch
from
da23be5 to
a82fced
Compare
| @@ -33,14 +33,17 @@ services: | |||
| volumes: | |||
| - ./:/app/:z | |||
There was a problem hiding this comment.
I think :z is linux specific, which gets ignored by Windows and macOS? Should we replace it with - ./app:/app to make it compatible across operating systems?
There was a problem hiding this comment.
Outside scope of PR - the volume is in reference todjango-app, not mlflow
mlflow/requirements.txt
Outdated
| @@ -0,0 +1,11 @@ | |||
| mlflow | |||
There was a problem hiding this comment.
Is mlflow along okay? Or do we need mlflow[genai]? See https://mlflow.org/docs/latest/llms/deployments/index.html
There was a problem hiding this comment.
Tracking server connects with chat endpoints, so I'm not sure if this is needed as it deploys as intended
There was a problem hiding this comment.
Do we have/need a ticket in the backlog to explore how to switch these OpenAI endpoints to our Redbox CoreAPI endpoint?
There was a problem hiding this comment.
Perhaps one to include in the backlog - definitely a good shout to explore connecting to the Redbox CoreAPI endpoint
Deploying to AWS is a separate ticket/piece of work, so let's add that to the backlog to track it. |
Context
Changes proposed in this pull request
Guidance to review
Relevant links
Things to check