DLISR: Function does not take argument(s) named device_id

[Ichunjo]

upscale = core.akarin.DLISR(clip.resize.Bicubic(format=vs.RGBS), 2, device_id=0)

vapoursynth.Error: DLISR: Function does not take argument(s) named device_id

It's also not recognised by vsrepo stubgen so I guess it's a different version of what the README says.

if relevant:

print(
    core.akarin.list_functions(), '\n',
    core.akarin.get_functions(), '\n',
)

DLISR(clip:clip; scale:int:opt)
Expr(clips:clip[]; expr:data[]; format:int:opt; opt:int:opt; boundary:int:opt)
Version()

 {'DLISR': 'clip:clip;scale:int:opt', 'Expr': 'clips:clip[];expr:data[];format:int:opt;opt:int:opt;boundary:int:opt;', 'Version': ''}

DLVFX doesn't seem to be compiled with this version
Version() {'expr_backend': b'llvm', 'expr_features': [b'x.property', b'sin', b'cos', b'%', b'N', b'X', b'Y', b'pi', b'width', b'height', b'trunc', b'round', b'floor', b'@', b'!', b'x[x,y]', b'x[x,y]:m'], 'version': b'v0.70-0-g85a4077'}

[AkarinVS]

I've committed support device_id to the repository but the code is not yet tested, so it's not released yet.

[AkarinVS]

I tested the code, and unfortunately, it doesn't work.

I also can't make the filter interoperate with other CUDA filters, so I've updated the docs to reflect these findings.

Based on experience with users of DLISR, the recommended way is this that you use a separate vspipe stage with DLISR alone (perhaps with bm3dcpu to clean up the input).
Also, as it's extremely slow, it's beneficial to save the y4m output as a file for future processing.