smc-finetuned-distilgpt2

SMC-Finetuned-DistilGPT2

This is a fine-tuned DistilGPT-2 model for generating trading signals based on Smart Money Concepts (SMC) for XAUUSD (gold) on a 15-minute timeframe. The model was trained on historical price data using simplified SMC rules (e.g., Break of Structure, Order Blocks, Fair Value Gaps, Liquidity Zones) to predict Buy/Sell/Hold signals.

Model Details

• Base Model: DistilGPT-2 (~82M parameters)
• Fine-Tuning: Used LoRA (Low-Rank Adaptation) for efficiency
• Dataset: Synthetic dataset derived from XAUUSD 15M OHLCV data with rule-based SMC signals
• Training: 3 epochs, 8-bit AdamW optimizer, effective batch size 16
• Input Format: Textual representation of 20-candle windows
• Output: Signals ("Buy", "Sell", "Hold")

The model uses simplified SMC rules; real-world performance requires backtesting. Trained on historical data; may need adaptation for live trading. Not financial advice—validate signals before use.