Sent140FederatedBenchmarkOptions

Options & Configuration · AiDotNet.Configuration

Configuration options for running the Sent140 LEAF federated benchmark suite.

For Beginners

You provide the train/test JSON files, and AiDotNet loads the per-user partitions, tokenizes the tweet text, and evaluates your model.

How It Works

Sent140 is a federated sentiment classification benchmark derived from tweets. LEAF stores the dataset as JSON where each “user” corresponds to a federated client.

Constructors

Constructor Summary
Sent140FederatedBenchmarkOptions()

Properties

Property Summary
LeafFederatedDatasetLoadOptions LoadOptions { get; set; } Gets or sets load options controlling how many users/clients are loaded.
int? MaxSamplesPerUser { get; set; } Gets or sets the maximum number of samples to use per user/client (null uses all available).
int? MaxSequenceLength { get; set; } Gets or sets the maximum token sequence length for each tweet after tokenization.
int? TokenizerTrainingSampleCount { get; set; } Gets or sets the maximum number of texts used to train a default WordPiece tokenizer when needed.
int? TokenizerVocabularySize { get; set; } Gets or sets the desired WordPiece vocabulary size when the model result does not already provide a tokenizer.
string TestFilePath { get; set; } Gets or sets the optional path to the Sent140 test split JSON file.
string TrainFilePath { get; set; } Gets or sets the path to the Sent140 train split JSON file.