Skip to main content

int8_ste

bitnet_quantize::layer

Function int8_ste 

Source 
pub fn int8_ste(input: &Tensor, scale: f32) -> Result<Tensor>
Expand description

Apply INT8 STE (quantize to [-127, 127]).

§Arguments

  • input - Input tensor
  • scale - Scale factor (typically AbsMax / 127)

§Errors

Returns error if tensor operations fail.