Update support of Megatron-LM PR 2

[pengdurice]

What does this PR do?

This PR adds additional details so that accelerate continues support using Megatron-LM as backend for training large scale LLMs.
In details:

1. initialize the model from hf checkpoint using meta device to avoid long time of loading weights which will not be used anyway.
2. introduce some additional parameters.
3. uses an initial_megatron_lm_checkpoint to load the model weights for fine-tuning
4. updates the user guide
5. add GLM4-moe model configs

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline,
  Pull Request section?
• Was this discussed/approved via a Github issue or the forum? Please add a link
  to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the
  documentation guidelines, and
  here are tips on formatting docstrings.
• Did you write any new necessary tests?

Who can review?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.

[pengdurice]

@SunMarc Hi, good Monday! This is another PR to update Megatron-LM with additional details for users to more easily use Megatron-LM. Would you please take a look when you got time, thanks!

[SunMarc]

Thanks a lot, just a minor nit !

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[SunMarc]

@bot /style

[github-actions]

Style bot fixed some files and pushed the changes.