pub fn read_f16_from_f32_buffer(buf: &GpuBuffer<f32>) -> GpuResult<Vec<f16>>Expand description
Read back a GpuBuffer<f32> and narrow each element to half::f16.
This is the counterpart of from_f16_slice_widening — it performs the
inverse narrowing conversion after the GPU computation has completed.
§Errors
Returns an error if the blocking read fails.