Trader Train

This trains LSTM, Transformer, and N-BEATS models on market data through the neural-trader package. You pick a model type and ticker symbol, and it runs training with confidence intervals, validation metrics, and prediction horizons up to 5 days out. The model comparison mode is handy since you can test all three architectures against the same symbol in one go. Results get stored to the trading-analysis namespace and fed into SONA for pattern learning. It's solid for quick ML backtesting without writing training loops yourself, though you're still dependent on neural-trader's implementation choices under the hood.