onnx-community

/

Florence-2-large

like 12

Follow

ONNX Community 605

Hello,
Thank you for you work!
I have a few questions regarding decoder model.

1. Original code's decoder has Florence2LearnedPositionalEmbedding module. There is no matching onnx file. Is it included to decoder's onnx?
2. Does decoder's onnx contain all 12 layers?
   Basically it is about entry and exit points of part exported to onnx. Original Florence2Decoder takes tokens ids on input, not embedding.
3. Did I get correctly that merged onnx should be used?

Thank you!