Skip to content

Sampler.md

Latest commit

 

History

History
16 lines (12 loc) · 938 Bytes

Sampler.md

File metadata and controls

16 lines (12 loc) · 938 Bytes
Raw

Sampler Tool

Accessing Sampler Tool

Navigate: Data Prep > Sample

Sampler Purpose

The sample tool allows you to create a random smaller subset of your dataset for testing and experimentation. This can be useful to randomly select a proportion of a dataset to manually label and compare to LLM results to check reliability. Additionally, it can be used to create a smaller dataset to experiment with LLM tools before scaling up to the full dataset.

Using the Sampler

  1. Select Dataset: Choose the dataset you want to sample from (accepted file types: csv, tsv, Excel, and Parquet)
  2. Select the Sampling Method: By rows will return the exact number of rows specified, while by percent will return the closest whole number of rows based on the proportion specified.
  3. Hit Ok: You will be prompted to save the sampled dataset as a csv. The default name is the original dataset name appended with "_sample".