adds fine tuning notebook and sample datasets

[justin-cechmanek]

No description provided.

[tylerhutcherson]

Left a few feedback items, nice work justin!

[tylerhutcherson]

Let's put this dataset in the public S3 bucket redis-ai-resources and then use wget or curl to download it. You will see folders in their for some of the others. We are trying to move more in this direction

[tylerhutcherson]

Same - let's move to S3 similar to above

[tylerhutcherson]

Nice! This is looking good. A few recommendations:

• Let's downplay the semantic caching use case specifics a bit more?
• Add hyper links to some of the relevant fine tuning papers when possible (contrastive loss, etc) or alternative techniques that could apply to other use cases (leaning more general)
• Potentially reduce the number of plots to just a few
• Possible to use the RedisVL huggingface vectorizer class to serve the embeddings here (just to insert some value prop?)

[justin-cechmanek]

yes to to first 3 points.
Not sure we can use our huggingface vectorizer here as it takes a string model name and pulls existing models from huggingface hub. I don't see a way to insert a local finetuned model. We could wrap it in our Custom vectorizer if we want to use it with our cache