mad-lab-qwen3-1.7b-sentiment
A QLoRA fine-tune of Qwen3-1.7B for financial sentiment analysis, trained on
earnings reports, SEC filings, and financial news to produce trading signal
classifications.
Intended Use
Classifying sentiment from financial text — earnings call transcripts, 10-K/10-Q
filings, analyst reports, and financial news — into actionable directional signals.
Training Details
- Base model: Qwen/Qwen3-1.7B
- Fine-tune method: QLoRA 4-bit (NF4 + double quant), LoRA r=32, alpha=64
- Trained on: GTX 1070 (~1.5-2h)
- Domain: Sentiment signals (news/filing sentiment → bullish/bearish/neutral)
- Targets: RSI/MACD/EMA-aligned sentiment labels from earnings reports and filings
