StackOverflowFederatedBenchmarkOptions
Options & Configuration · AiDotNet.Configuration
Configuration options for running the StackOverflow federated benchmark suite.
For Beginners
You provide the train/test JSON files, and AiDotNet loads the per-user partitions, builds a vocabulary (with safe defaults), and evaluates your model on a standardized next-token task.
How It Works
StackOverflow is a large-scale federated text benchmark commonly used for next-token prediction. This v1 suite expects a LEAF-style JSON container with per-user token sequences.
Constructors
| Constructor | Summary |
|---|---|
StackOverflowFederatedBenchmarkOptions() |
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? MaxVocabularySize { get; set; } |
Gets or sets the maximum vocabulary size used for token-to-ID mapping. |
int? SequenceLength { get; set; } |
Gets or sets the fixed token sequence length used as model input. |
int? VocabularyTrainingSampleCount { get; set; } |
Gets or sets the maximum number of sequences used to build the default vocabulary. |
string TestFilePath { get; set; } |
Gets or sets the optional path to the StackOverflow test split JSON file. |
string TrainFilePath { get; set; } |
Gets or sets the path to the StackOverflow train split JSON file. |