Skip to content

Datasets

A dataset is what happened in one measurement (or one published computation). It is not “the S21 trace” — the trace is a child. The dataset holds:

  • One or more files (Touchstone .s1p–.s4p / .snp, CSV, images)
  • Traces parsed from those files (S11, S21, …)
  • Parameters — key/value test conditions (lot=15, temp=25, instrument=182)
  • Tags — team-scoped labels (needs-review, golden)

How you meet a dataset

The project page leads with notebooks. Every new dataset gets a notebook automatically so you can plot it immediately. Open the dataset itself when you need files, parameters, or tags.

In the SDK:

ds = cs.projects["cable-qual"].datasets["sweep-25c"]
s21 = ds.traces["S21"]
print(ds.params)

Bracket lookup is exact. Glob works on names (datasets["sweep-*"]). Parameter filters are usually better than name globs:

matches = cs.projects["cable-qual"].datasets.filter(cs.params.lot == "15")

Parameter values are strings. Filter with "15", not 15.

Files and traces

Upload a Touchstone file onto the dataset and CloudSprite parses standard S-parameter traces. CSV becomes x/y traces from columns. You work with traces in notebooks and in the SDK; you download the original file when you need the vendor’s raw capture.

Large files upload in chunks. Wait for parsing to finish before you assume traces exist.

Creating datasets

  • App: from a project, create a dataset and upload files.
  • SDK: upload onto an existing dataset with ds.upload("part.s2p"), or cs.publish(...) to create a new dataset from computed traces (see Compute and publish results).

Do not treat a published result as a second-class object. It is a dataset in the same project, with its own notebook, parameters, and (if you opted in) the script that produced it.