New learning path: Deploy Arcee AFM-4.5B on AWS Graviton4

[juliensimon]

[X] I have reviewed Create a Learning Path
[X ] I have checked my contribution for confidential information
By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of the Creative Commons Attribution 4.0 International License.

This is the Arcee AI / AWS Graviton learning path discussed with Ayasha Ali and Tobias McBride.
I also have a GCP Axion version, which I'll submit once this PR has been reviewed. A lot of the content is very similar.

IMPORTANT NOTE: The AFM-4.5B model is not yet publicly available on Hugging Face. It will be in a few weeks.

This PR shouldn't be merged until then.

A couple of TO DOs before merging:

• Update the HF repository name
• Update perplexity numbers

[jasonrandrews]

Hi @juliensimon

Thanks for the great Learning Path. I'm stuck on the command:

huggingface-cli download arcee-ai/afm-4.5B --local-dir models/afm-4-5b

Can you confirm the model is public and available to download?

[juliensimon]

Hi @jasonrandrews

As explained in the PR, the model isn't public yet. You can use arcee-ai/Llama-3.1-SuperNova-Lite, as the process is exactly the same. Thank you.

[jasonrandrews]

@juliensimon Thank you, sorry I missed that. We will complete the editorial review and I will recheck the tech steps when the new model is ready and update the command.

[juliensimon]

Hi @jasonrandrews,

The model will be on Hugging Face in the next few days. I'll keep you posted.

Any feedback on the editorial review so far?

[jasonrandrews]

Hi @juliensimon,
In final review now so we are ready when the model is ready. Looking good, I don't foresee any issues to publish.

[jasonrandrews]

Hi @juliensimon
The review is done except a few small things like model available and the TODO items in the data. Please review everything on the main branch. It's currently in draft mode so you can clone the main branch and use hugo server -D to show the draft articles.

[juliensimon]

Hi @jasonrandrews,

ok, thank you. I'll take one last look, and I'll resolve the last few TODOs as soon as the model is on Hugging Face. Should I create a new PR once the article is final?

I'll also create another PR for the GCP version, which will be identical, except for instance setup, of course.

[jasonrandrews]

@juliensimon
Yes, you can create a new PR with any remaining updates.