Cisco Pipeline

[abaranov25]

Resolve #663
Adding a Cisco pipeline to Orion

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you all sign our Contributor License Agreement before we can accept your contribution.
1 out of 2 committers have signed the CLA.

✅ abaranov25
❌ AllenBaranov

---

AllenBaranov seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[sarahmish]

I prefer to have these values similar to the other pipelines, i.e.

Suggested change

	"window_size": 30720,
	"step_size": 16
	"window_size": 250,
	"step_size": 1

Is there a reason for this value?

[abaranov25]

Yes. 30720 fine samples are required to fully populate the coarse context for cisco. With an aggregation factor of 60 and 512 coarse points, we need 512*60=30720 points. We could use 250 points by default, but there would only be 4 long term context points, all of which are reflected in the fine context. It would turn the model into (basically) a timesfm clone.

[sarahmish]

if window_size and window_step_size are specified, then you can remove window_size_portion and window_step_size_portion

[sarahmish]

Is series a raw forecast? 1D array of what exactly?

Suggested change

	series: 1D numpy array
	agg_factor: Aggregation factor (default 60 = minute->hour)
	series (ndarray):
	1D numpy array
	agg_factor (int):
	Aggregation factor (default 60 = minute->hour)

[sarahmish]

This should be described instead of returning variable names