v0.18.0

What's Changed

• Add support for fine-tuning and files using the Azure API. by @sorinsuciu-msft in #80
• feat(wandb): sanitize name by @borisdayma in #77
• Add customer endpoints (#126) by @hallacy in #87

Full Changelog: v0.16.0...v0.18.0

v0.16.0

What's Changed

• fix(wandb): no results if not succeeded (#74)
• [lint] standardize isort and black in pyproject.toml (#81)
• Add support for edit call
• Add version bump (0.16.0)

New Contributors

• @tomerk made their first contribution in #81

v0.15.0

What's Changed

• Add a suffix arg to fine_tunes.create that allows you to customize the generated model name. See #75. For example, openai api fine_tunes.create -t test.jsonl -m ada --suffix "custom model name" could generate a model with the name ada:ft-your-org:custom-model-name-2022-02-15-04-21-04.

New Contributors

• @tomerk made their first contribution in #75

v0.14.0

What's Changed

• Add entry_points by @BastianZim. Makes some changes to setup so that the package can be easily installed on Windows
• Add documentation for usage with Azure endpoints by @sorinsuciu-msft
• feat: openai wandb sync by @borisdayma. Adds Wandb integration. Try openai wandb sync as an example

New Contributors

• @BastianZim made their first contribution in #52
• @borisdayma made their first contribution in #64

v0.13.0

• Adds new example notebook demonstrating embeddings for recommendations
• Updates other embedding examples with new engine names
• Adds new functions to embedding_utils.py

v0.12.0

• Add an option to use Azure endpoints for the /completions & /search operations

v0.11.6

Changes in this version:

• Deprecating openai.Engine.embeddings. Instead, please use openai.Embedding.create
• Uses the new embedding engine names in all examples
• Delete the ca certificates that aren't needed anymore
• Makes all openai exceptions pickle-able

v0.11.5

• remove packing flags from fine_tunes.create
• remove recommended batch sizes from data prep tool (default batch sizes now set server-side)

v0.11.4

• Answers and Classification endpoints now have their own prepare_data functions, similar to search
• Moved the utils file used in the embeddings examples to embeddings_utils. This will allow the code to be imported
• Changed the default api_prefix to be "" and instead moved that logic into init.py. You might need to update any local OPENAI_API_BASE that you've set.

v0.11.3

• Hotfix to deal with missing import