Open Documentation
Data & Evaluation
Before a fine-tune can work, the data has to be right. This section covers deduplication, dataset validation, and the statistics behind trusting a benchmark result, the checks worth running before you spend a GPU hour.
AllData & EvaluationPost-Training TechniquesModel Architecture & InternalsDeployment & Infrastructure
Data & Evaluation·Deep Dive
Power Analysis for Benchmark Design | SR Cookbooks
Calculate statistical power and minimum detectable accuracy gaps for LLM benchmarks using Python to avoid reporting sampling noise as signal.
PythonEvaluationsStatisticsBenchmarksData Preprocessing
Data & Evaluation·Deep Dive
Deduplication at Scale: MinHash & LSH in Python | SR Cookbooks
Implement MinHash and Locality-Sensitive Hashing (LSH) in pure Python to eliminate the quadratic bottleneck of near-duplicate detection for ML dataset curation.
PythonAlgorithmsData PreprocessingMinHashLSH
Data & Evaluation·Deep Dive
How to Check a Fine-Tuning Dataset Before You Train | SR Cookbooks
Avoid common fine-tuning failures. Learn how to validate chat templates, prevent silent truncation, verify loss masking, and check for data leakage before spending GPU hours.
DatasetsFine-TuningEvaluation