Closed juhoinkinen closed 11 months ago
OpenAI client has changed since running this notebook (v1.0 was released 6.11.2023, and now the latest is already v1.3.7).
Also the babbage model has been deprecated and will shutdown 4.1.2024, so I updated to use babbage-002 base model as recommended.
For my run the results got worse (overall similarity 0.7886 -> 0.7216) but at this point I think it does not matter so much.
BTW, at openai.com one can now see logs and training loss of the fine-tune run:
OpenAI client has changed since running this notebook (v1.0 was released 6.11.2023, and now the latest is already v1.3.7).
Also the babbage model has been deprecated and will shutdown 4.1.2024, so I updated to use babbage-002 base model as recommended.
For my run the results got worse (overall similarity 0.7886 -> 0.7216) but at this point I think it does not matter so much.
BTW, at openai.com one can now see logs and training loss of the fine-tune run: