Fix Feast's Jupyter notebook code

[samhita-alla]

Signed-off-by: Samhita Alla aallasamhita@gmail.com

• Jupyter notebook works now; included minio environment variables
• Bumped Feast version to 0.15 (will remove proto-plus dependency once 0.16 is out)

[kumare3]

Does this work? How can it get the right thing

[samhita-alla]

I didn't get you. Why shouldn't it?

[samhita-alla]

If you mean to ask if the output is a FeatureStore data class with all the methods within it accessible, then no, it isn't precisely so. Hence, what I am doing is, accessing the config and creating a FeatureStore out of it.

import os

from feast_workflow import predict, FEAST_FEATURES
from feast_dataobjects import FeatureStore

os.environ["FLYTE_AWS_ENDPOINT"] = os.environ["FEAST_S3_ENDPOINT_URL"] = "http://localhost:30084/"
os.environ["FLYTE_AWS_ACCESS_KEY_ID"] = os.environ["AWS_ACCESS_KEY_ID"] = "minio"
os.environ["FLYTE_AWS_SECRET_ACCESS_KEY"] = os.environ["AWS_SECRET_ACCESS_KEY"] = "miniostorage"

inference_point = FeatureStore(fs.config).get_online_features(FEAST_FEATURES, [{"Hospital Number": "533738"}])
inference_point

Does this not look okay? I tried running the code with raw_outputs and this is what I see:

AttributeError: 'FlyteWorkflowExecution' object has no attribute 'raw_outputs'

[kumare3]

Ohh ya I had hacked that method, I think there is a better method now

[pingsutw]

@kumare3 we haven't merged that PR yet, so we can't get "raw_outputs" now.
flyteorg/flytekit#675
I think that PR is ready to be merged after resolving the merge conflict.

[samhita-alla]

@pingsutw, can you please merge flyteorg/flytekit#675? I think @kumare3 wants you to finalize, clean up the code, and merge the PR.

[samhita-alla]

@kumare3, code is working now. Can we merge this PR?