Hamilton Open Source

<@U05H6BRN00G> I would guess that it's checking the input being passed in via `inputs=` and not thinking it matches the annotation for `input_image`.

Which could be a bug in how we're comparing.

Shouldn't be hard to repro. Mind sharing the library version?

<@U03AV0GG20H> basic repro here:

```from PIL import Image

def foo(bar: Image) -&gt; dict:
    return {'a' : 1}

from hamilton.ad_hoc_utils import create_temporary_module

mod = create_temporary_module(foo)

from hamilton.driver import Driver



if __name__ == "__main__":
    img = Image.open("test.png")
    dr = Driver({}, mod)
    print(dr.raw_execute(final_vars=['foo'], inputs={"bar" : img}))```


If you replace the type annotation with `ImageFile.ImageFile` (:/) it works!

```from PIL import Image, ImageFile

def foo(bar: ImageFile.ImageFile) -&gt; dict:
    return {'a' : 1}

from hamilton.ad_hoc_utils import create_temporary_module

mod = create_temporary_module(foo)

from hamilton.driver import Driver



if __name__ == "__main__":
    img = Image.open("test.png")
    dr = Driver({}, mod)
    print(dr.raw_execute(final_vars=['foo'], inputs={"bar" : img}))```

Also this is using the newer pillow (I think `.read()` has been deprecated) If I were you, I’d just do in a constants file:

```ImageType = ImageFile.ImageFile```
But I also suggest validating the type for your version of the pillow library or updating it

So <@U05H6BRN00G> I think the work around is trying change the type annotation as Elijah suggested, otherwise will dig into the discrepancy in behavior - I created <https://github.com/DAGWorks-Inc/hamilton/issues/223> to track.