feat: add build_model function

[LMMilliken]

This PR adds a build_model using features added in pr 336.

---

• This PR references an open issue
• I have added a line about this change to CHANGELOG

[bwanglzu]

minor comments

[gmastrapas]

1. Also update changelog
2. Include a unit test for this method

[guenthermi]

added some comments

[bwanglzu]

no need to change, but a better way is

Suggested change

	@pytest.mark.parametrize(
	'descriptor, select_model, is_onnx, expect_error',
	[
	('bert-base-cased', None, False, None),
	('bert-base-cased', None, True, None),
	('openai/clip-vit-base-patch16', 'clip-text', False, None),
	('openai/clip-vit-base-patch16', 'clip-vision', False, None),
	('openai/clip-vit-base-patch16', None, False, SelectModelRequired),
	('MADE UP MODEL', None, False, ValueError),
	],
	)
	@pytest.mark.parameterize('is_onnx', [True, False])
	@pytest.mark.parametrize(
	'descriptor, select_model, expect_error',
	[
	('bert-base-cased', None, None),
	('openai/clip-vit-base-patch16', 'clip-text', None),
	('openai/clip-vit-base-patch16', 'clip-vision', None),
	('openai/clip-vit-base-patch16', None, SelectModelRequired),
	('MADE UP MODEL', None, ValueError),
	],
	)

[LMMilliken]

The more you know!

[bwanglzu]

LGTM!

[gmastrapas]

One minor thing

[LMMilliken]

.

[gmastrapas]

LGTM!

[guenthermi]

I added too comments, rather minor

[guenthermi]

Here is an annotation of the return type missing

[guenthermi]

if build_model returns nothing it would actually pass. Testing something with the model would be better. Maybe you can just check if you can pass an input and the result is a vector of the expected shape. If this is too much, you could at least check if the type of model is correct and not None

[guenthermi]

LGTM, but I think documentation for it is still missing, but you can do this in a separate PR.