| Topic | Posts | Latest |
|---|
| Upgrade Trapswhat changes, often silently, when you upgrade pandas, Polars, Python, scikit-learn or PyTorch | 7 | |
| Validation and Leakagepreprocessing leaks, split mistakes, and metrics that look better than the model is | 9 | |
| Retrieval and RAGchunking, embeddings, BM25, reranking and thresholds, measured on the same question set | 9 | |
| LLM Inferencetokens, decoding, KV cache, precision and padding, run on a small local model | 12 | |
| Time Series ForecastingLSTM input shapes, time-ordered validation, and multi-step forecasts against simple baselines | 4 | |
| Distributions and Basic Statisticshistograms, bins, KDE, mean vs median and normality tests | 6 | |
