[BUG] Tensor metadata not updated when calling astype()

[fnands]

Bug description

Hey, I have run into something odd with Tensors.

The tensor astype() method: https://docs.modular.com/mojo/stdlib/tensor/tensor/Tensor#astype

Says:

Copy the Tensor with elements cast to the new type.

However when I do this the metadata of the tensor seems to still be that of the old tensor type.

The values in the tensor themselves are fine, but when looking at the bytecount and tensorspec are called the new tensor seems to think it is still the old tensor.

Steps to reproduce

Minimal example:

from tensor import Tensor, rand

var tensor_uint8 = rand[DType.uint8](3, 3)
print("8 bit tensor byte count: ", tensor_uint8.bytecount())
var tensor_float32 = tensor_uint8.astype[DType.float32]()
print("32 bit tensor byte count: ", tensor_float32.bytecount())

print("")
print("8 bit tensor spec: ", tensor_uint8.spec())
print("32 bit tensor spec: ", tensor_float32.spec())

Yeilds:

8 bit tensor byte count:  9
32 bit tensor byte count:  9

8 bit tensor spec:  3x3xuint8
32 bit tensor spec:  3x3xuint8

System information

• What OS did you do install Mojo on ?
  Ubuntu 22.04.4 LTS
• Provide version information for Mojo by pasting the output of mojo -v
  mojo 24.3.0 (9882e19d)
• Provide Modular CLI version by pasting the output of modular -v
  modular 0.7.2 (d0adc668)

[JoeLoser]

This is fixed in a recent nightly ~1 week ago