pub fn read_f32(view: &TensorView<'_>, name: &str) -> Result<Vec<f32>>Expand description
Reads view’s raw bytes as f32, upcasting from F32/F16/BF16 as needed. Integer or
boolean source tensors (already-quantized inputs, masks, etc.) aren’t supported – callers
should pass those through unquantized rather than routing them through this function.