Remove _copy_tensor from usage

[rohan-varma]

Context

• _copy_tensor was not needed and was added due to a misunderstanding on how pytorch internally handles tensor reference counting. The thinking was that we're creating a linear layer and want to save the bias, so we just deep clone the bias before returning it. This however is inaccurate, and its sufficient to just return a reference to the bias. This is because torch will internally bump the refcount to the bias's storage by 2, and when the function goes out of scope, we'll still have a reference count of 1 due to the returned bias variable.
• This change already exists for weight.

Changelog

• Get rid of _copy_tensor

Test plan

• CI

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/torchtune/633

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit a38d17c with merge base 2940941 ():
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[kartikayk]

Thanks for revisiting this!