ONNX
#2
by
cnmoro
- opened
Is it possible to convert this model to onnx and use it?
I've been looking for a small enough model that would be a good fit to be actually embedded into a pypi package (torch library automatically downloads nvidia dependencies and is huge)
Surprisingly, this 100M parameter model is amazing for it's size. Awesome!
I have tried to convert this to onnx for about two weeks. I used spox and it work eventually for model size below 2GB.
I have trouble converting the model above 2GB to onnx, does anyone has any recommendation for this?