Fix prev

crates/cubecl-core/src/frontend/element/float/tensor_float.rs

@@ -1,9 +1,12 @@
+#![allow(clippy::transmute_int_to_float)] // Not yet stable in previous version. To be removed when
+                                          // prev=1.83.
+
 use bytemuck::{Pod, Zeroable};
 use core::ops::{Add, AddAssign, Div, DivAssign, Mul, MulAssign, Sub, SubAssign};
 use half::f16;
 use num_traits::{NumCast, ToPrimitive};
 use serde::Serialize;
-use std::num::NonZero;
+use std::{mem::transmute, num::NonZero};
 
 use crate::{
     ir::{Elem, FloatKind, Item},
@@ -34,7 +37,7 @@ impl tf32 {
     #[inline]
     #[must_use]
     pub const fn from_bits(bits: u32) -> tf32 {
-        tf32(f32::from_bits(bits))
+        tf32(unsafe { transmute::<u32, f32>(bits) })
     }
 
     /// Constructs a [`tf32`] value from a 32-bit floating point value.
@@ -64,7 +67,7 @@ impl tf32 {
     #[inline]
     #[must_use]
     pub const fn to_bits(self) -> u32 {
-        self.0.to_bits()
+        unsafe { transmute(self.0) }
     }
 
     /// Converts a [`tf32`] value into an [`f32`] value.