[workflow] Add Phi-4 Finetuning Example

[rajithkrishnegowda]

This PR adds Phi-4 Finetuning example using Workflow API.

[MasterSkepticista]

1. I have added a Federated Learning (FL) tutorial that allows users to input and submit the number of rounds for federated learning. This feature enhances interactivity and helps users better understand the iterative nature of federated learning.

• For (1), my recommendation is to remove gradio and the interactive element (if this existed for older notebooks, I suggest updating them too). A notebook interface is interactive enough. IMO adding GUI is out-of-place for a fine-tuning example. Furthermore, we really want to reduce the amount of version maintenance we do.
• For (3), my suggestion was to actually not install openfl in the notebook. It is expected that the user has already installed OpenFL by the time they clone the repository. It is perhaps the first thing they are expected to do. In this (and all other) notebooks, a markdown cell can point the user to installation page on official docs and then the step to activate experimental APIs. Installing any packages specific to the notebook is fine, though.

[psfoley]

1. I have added a Federated Learning (FL) tutorial that allows users to input and submit the number of rounds for federated learning. This feature enhances interactivity and helps users better understand the iterative nature of federated learning.

• For (1), my recommendation is to remove gradio and the interactive element (if this existed for older notebooks, I suggest updating them too). A notebook interface is interactive enough. IMO adding GUI is out-of-place for a fine-tuning example. Furthermore, we really want to reduce the amount of version maintenance we do.
• For (3), my suggestion was to actually not install openfl in the notebook. It is expected that the user has already installed OpenFL by the time they clone the repository. It is perhaps the first thing they are expected to do. In this (and all other) notebooks, a markdown cell can point the user to installation page on official docs and then the step to activate experimental APIs. Installing any packages specific to the notebook is fine, though.

1. Agree on this point. I'm personally a fan of gradio - and it would be useful to show in potentially other dedicated notebooks how other higher level interfaces can be built on top of the Workflow API, but in this notebook we want the focus to be LLM usage.
2. OpenFL indeed should be installed ahead of time. workflow_interface_requirements.txt is only kept as separate for now because this is an experimental package. These will eventually be merged into the requirements of setup.py once the experimental label for the Workflow API is dropped. Installing packages needed for a particular notebook example is standard practice.

[rajithkrishnegowda]

1. I have added a Federated Learning (FL) tutorial that allows users to input and submit the number of rounds for federated learning. This feature enhances interactivity and helps users better understand the iterative nature of federated learning.

• For (1), my recommendation is to remove gradio and the interactive element (if this existed for older notebooks, I suggest updating them too). A notebook interface is interactive enough. IMO adding GUI is out-of-place for a fine-tuning example. Furthermore, we really want to reduce the amount of version maintenance we do.
• For (3), my suggestion was to actually not install openfl in the notebook. It is expected that the user has already installed OpenFL by the time they clone the repository. It is perhaps the first thing they are expected to do. In this (and all other) notebooks, a markdown cell can point the user to installation page on official docs and then the step to activate experimental APIs. Installing any packages specific to the notebook is fine, though.

changes made in 6ebb95e

[MasterSkepticista]

Almost there. A few polish tips:

1. Heading styles inconsistent across the notebook. Suggest using h1 for title (i.e. single # in markdown) and h2 for sections, h3 for sub-sections. Note that I don't mean to enforce using h1-h2-h3. Some notebooks are also good with just h1 and h3.
2. Update PR description (since now not using gradio)
3. Re-run the notebook and include cell outputs. As mentioned in previous notebook PRs, having cell outputs is useful for the user. This helps a user guess whether the notebook is worth their time investment very quickly.

[rahulga1]

Almost there. A few polish tips:

1. Heading styles inconsistent across the notebook. Suggest using h1 for title (i.e. single # in markdown) and h2 for sections, h3 for sub-sections. Note that I don't mean to enforce using h1-h2-h3. Some notebooks are also good with just h1 and h3.
2. Update PR description (since now not using gradio)
3. Re-run the notebook and include cell outputs. As mentioned in previous notebook PRs, having cell outputs is useful for the user. This helps a user guess whether the notebook is worth their time investment very quickly.

@MasterSkepticista I have a different view on keeping the cell output in notebook. It hampers the readability, these examples are there for reference purpose to demonstrate capability.
Anyways we are in process of removing non-essential examples out of this repo and the one which we keep will be in included in CI, to ensure new changes are not breaking examples.

[rajithkrishnegowda]

Almost there. A few polish tips:

1. Heading styles inconsistent across the notebook. Suggest using h1 for title (i.e. single # in markdown) and h2 for sections, h3 for sub-sections. Note that I don't mean to enforce using h1-h2-h3. Some notebooks are also good with just h1 and h3.
2. Update PR description (since now not using gradio)
3. Re-run the notebook and include cell outputs. As mentioned in previous notebook PRs, having cell outputs is useful for the user. This helps a user guess whether the notebook is worth their time investment very quickly.

1. changes made in f785846
2. updated PR description

[kta-intel]

Looking good @rajithkrishnegowda, just some minor questions/comments

[porteratzo]

Looks good to me

[MasterSkepticista]

Thanks for the updates @rajithkrishnegowda , it looks polished now!

[kta-intel]

thanks for the changes, LGTM!