Ludwig is an project opensourced by uber in 2019, it can help to train model easily and with the new release published, 0.8.x, it supports the LLM required capacities, like the SFT, so I think build on top of Ludwig can save us a lot of time.
Discussed offline with @ zhijiang, considering we have little people working on this project right now, and we should focus more on the deployment in kubernetes, so generally it's ok.
Ludwig is an project opensourced by uber in 2019, it can help to train model easily and with the new release published, 0.8.x, it supports the LLM required capacities, like the SFT, so I think build on top of Ludwig can save us a lot of time.
Discussed offline with @ zhijiang, considering we have little people working on this project right now, and we should focus more on the deployment in kubernetes, so generally it's ok.