Closed HarryWu99 closed 4 months ago
https://github.com/opengear-project/GEAR/blob/79ad3fcdb528fceaf605923479fe14fdf3953ffd/TrueCompressionLlaMA/models/TrueCompressFunction.py#L128
error = input - torch.round((input - min) / step)
but it should be error = input - (torch.round((input - min) / step) * step + min). So that the error is the quantization residuals.
error = input - (torch.round((input - min) / step) * step + min)
error
Thanks for reminding us. We have fixed this error in our new version with fused quantized operator support. Will upload soon.
https://github.com/opengear-project/GEAR/blob/79ad3fcdb528fceaf605923479fe14fdf3953ffd/TrueCompressionLlaMA/models/TrueCompressFunction.py#L128
error = input - torch.round((input - min) / step)
but it should be
error = input - (torch.round((input - min) / step) * step + min)
. So that theerror
is the quantization residuals.