Pass Inputs to Outputs when Wired

[yeldarby]

Description

This fixes a bug where if you add one of the input parameters as an output the default value doesn't make its way through. This is useful when wanting to display input values (like a hard-codes string) in your rendering code.

Type of change

• Bug fix (non-breaking change which fixes an issue)

How has this change been tested, please provide a testcase or example of how you tested the change?

We implemented this to use in @josephofiowa's demo last week & have now tested it several times end to end with new workflows (installing inference from this branch).

On top of that, we want to create 0.11.1 release, so:

• checked that CI still 🟢
• Conducted tests on target platforms:
  • JETSON 4.5: image builds
  • JETSON 4.6.1 tests 🟢 https://github.com/roboflow/inference/actions/runs/9203845879
  • JETSON 5.1.1 - local build 🟢
  • NVIDIA GPU tests 🟢 https://github.com/roboflow/inference/actions/runs/9203849682

Any specific deployment considerations

No

Docs

• Shouldn't be needed; fixes a bug to make it work the way it'd be expected to

[PawelPeczek-Roboflow]

We are impacted by new release of setuptools that happened yesterday. https://pypi.org/project/setuptools/#history

Root cause:

• we use rf-clip heavily relying on packaging which is legacy and no longer supported
• https://setuptools.pypa.io/en/stable/pkg_resources.html

Impact:

• everything installs fine with older versions of setuptools (namely <70.0.0), but people using newer environments will get impacted sooner or later
• I am pinning setuptools<70.0.0 in requirements, but this is not the ultimate solution for the problem (as we need to align with newer setuptools, which will be probably the direction taken by growing number of packages over time)
• opened an issue in original repo: from pkg_resources import packaging seems not work with setuptools>=70.0.0 openai/CLIP#446, if that does not help, we will port on our own
• when it comes to releasing that change - we shall do it, but there will be need for tmp solutions in at least some cases, which will be enforcing additional pip install setuptools<70.0.0, as we cannot undo problems for legacy versions

[PawelPeczek-Roboflow]

We should probably distribute note of the following (or similar) content to people using inference

Using setuptools in version 70.0.0 and above breaks usage of Clip and YoloWorld models. That impacts historical version of inference package installed in python environments with newest setuptools.
We will release change pinning setuptools version into compatible one soon, working on the solution for that constraint not to be needed.

As a solution for now, we recommend enforcing setuptools<70.0.0 in all environments using inference:
`pip install setuptools>=65.5.1,<70.0.0`

[PawelPeczek-Roboflow]

@yeldarby I will make a release from that branch once u confirm u are fine with this

[yeldarby]

Yep looks good (I think you have to be the one to actually ✅ because I’m the one who submitted the PR though).

[PawelPeczek-Roboflow]

ok, merging and I will release as 0.11.1 later afternoon

[PawelPeczek-Roboflow]

Preparing release from this branch:
JETSON 4.5: image builds
JETSON 4.6.1 tests 🟢 https://github.com/roboflow/inference/actions/runs/9203845879
JETSON 5.1.1 - local build 🟢
NVIDIA GPU tests 🟢 https://github.com/roboflow/inference/actions/runs/9203849682