Fill in architecture compatibility data

This file is incomplete, I think. Check that the compatibility data is correct and fill in any gaps.

Compatibility means that any model of that architecture can be converted to the compatible platforms.

Can we call pytorch/onnx/ncnn frameworks and not architectures please

I called them platforms.

export type Platform = 'pytorch' | 'onnx' | 'ncnn';

Archs are e.g. ESRGAN. Or in general, the architecture field of models.

You used architecture in the issue post

Fixed.

@theflyingzamboni

With #235, we just need to test EDSR (SRResNet), which we currently can't do because chainner doesn't support it.

Also, we (speak @theflyingzamboni) are currently working on NCNN for SPSR. This should be possible, but there seems to be an issue with our conversion code. See our private conversation on discord.

Navigation Menu