ziyanxzy

this issue in einsum node: output: "/Einsum_output_0" name: "/Einsum" op_type: "Einsum" attribute { name: "equation" type: STRING s: "nhwpqc->nchpwq" }

> Could you post more detailed information? Screenshot of failure and the model including weight(if possible) would be helpful. i provide the xml model, and you can compile this model...

> we've tested fastvideo on window's WSL. Have not tried compiling VSA, but don't see why it wouldn't work and vsa only support hopper arch? like h100, can not support...

> VSA is supported for non hopper arch. It will just be using a triton implementation instead of thunderkittens thx a lot . but st attention Just for h100 and...

but he group shape in this model is always 512 not 1. it is not support depth-separable convolution?

i have the same problem. when i choose dataset "piqa" limit 5 ,it can work. but if i use limit10/20/50 and more cases, it can not. 

 > > @nazneenn Maybe I can search for debug logs if exists to help you understand this problem a bit more? > > @MaksimDanilov, you may try any debugger...