Vector dtype recipe update for int8

[justin-cechmanek]

No description provided.

[tylerhutcherson]

Realizing this actually depends on the upcoming release of redisvl. So we can hold it in a PR and merge / test once the package is available (no big deal). Make sure to update the README and colab links since you changed notebook name (which I agree with).

Also is it always practical to inflate your integers using a scaled method before casting to int8? Aren't there models from hugging face or cohere we can showcase here for int8?

[justin-cechmanek]

There some non-scaled huggingface models, but the first I found still have typical values between (-5.0, +5) so using those raw still has a lot of rounding error when cast to int.

[justin-cechmanek]

In our test environment we're using the docker image redis-stack-server:latest, which still is on Redis 7.4. @tylerhutcherson we can update the test env, or wait until the latest docker image uses Redis 8.0 for int8 support. Thoughts?

[tylerhutcherson]

@justin-cechmanek let's update the CI to use the redis:8.0-M03 docker image. Good catch. Assuming that fixes the issue here, let's try to get that updated and we can merge. Tag me whenever ready!