Spaces:

---

sergiopaniego

/

vlm_object_understanding

like 17

Running on Zero

Added an extra output that shows inference time for both models. I removed the @GPU annotation on the 'detect' function because it was causing a ~400ms hit to the first model's inference time (presumably because of ZeroGPU initialization that the second model is able to take advantage of). With the annotation removed both models get the performance hit resulting in a more apples-to-apples comparison.

Here's what it looks like: