HellaswagDataLoaderOptions
Options & Configuration · AiDotNet.Data.Text.Benchmarks
Configuration options for the HellaSwag commonsense NLI benchmark (Zellers et al. 2019).
How It Works
HellaSwag is a 4-way multiple-choice commonsense reasoning benchmark where each example presents a context sentence and 4 possible endings; only one is the natural continuation. ≈ 39,905 train / 10,042 val problems. Adversarially filtered against early LMs — remains one of the standard 0-shot LM-eval benchmarks today.
Constructors
| Constructor | Summary |
|---|---|
HellaswagDataLoaderOptions() |
Properties
| Property | Summary |
|---|---|
DatasetSplit Split { get; set; } |
Dataset split to load. |
bool AutoDownload { get; set; } |
Automatically download if not present. |
int MaxSequenceLength { get; set; } |
Maximum sequence length per (context + ending) pair. |
int VocabularySize { get; set; } |
Maximum vocabulary size. |
int? MaxSamples { get; set; } |
Optional maximum number of samples to load. |
string DataPath { get; set; } |
Root data path. |