chore: remove fine tuning support

[hrnn]

We are discontinuing support for fine-tuning, as va.landing.ai no longer offers this feature. However, we will continue to support custom object detection model training on LandingLens (app.landing.ai).

Warning

This is a breaking change for old chats that used fine-tuned models.

[Dayof]

old chats might be calling the functions with fine_tune_id, isn't it better to be backward compatible and check if fine_tune_id is set, print a warning saying that this arg is deprecated and add the deployment_id as a separate arg?

[CamiloInx]

old chats might be calling the functions with fine_tune_id, isn't it better to be backward compatible and check if fine_tune_id is set, print a warning saying that this arg is deprecated and add the deployment_id as a separate arg?

Agree. We should probably don't modify the fine_tune_id param, and only use it for the custom_od from now on.

[Dayof]

can you add a test to make sure that fine_tune_id still acessible

[Dayof]

check that fine tune id is still accessible to all the tests that used to call the fine tune logic, not only one

[camiloaz]

i don't understand these changes very well. we are removing the code that processes fine_tune_id but we don't remove the fine_tune_id parameter from the function signatures. which means that the code will receive the parameter, raise a deprecation warning but still will not use the parameter for anything. i don't think this makes sense.

why are we concerned with making this change backwards compatible? we can bump a major version if we want to, people using the vision-agent can decide whether to upgrade the version or not.

[hrnn]

Hi @Dayof @CamiloInx
Recently we agreed with @dillonalaird to don't make these changes backward compatible.

[dillonalaird]

LGTM