Fetching metadata from the HF Docker repository...
-
Add predict_api endpoint (no plot, faster) Running
-
Add align_to_hour parameter for data mode selection
-
Use Binance public REST API with multiple endpoints fallback
-
Use CPU PyTorch and ccxt for data (bypass Binance geo-restriction)
-
Fix Python version to 3.10, upgrade Gradio to 5.9.1
-
Deploy Kronos BTC Forecast API
-
Improve down-prediction accuracy: 44% -> 63%, overall 55% -> 62%
xianqiu
Enhanced training v4:
- Weighted sampling (1.5x for down samples)
- Imbalanced dataset (55% down / 45% up)
- 8 epochs with lower LR (3e-6)
- Converged after 2 iterations
Metrics:
- Direction accuracy: 55.1% -> 62.3%
- Down-prediction accuracy: 44.2% -> 62.6%
- Up-prediction accuracy: 65.9% -> 62.0% (slight decrease, better balanced)
-
Update model with balanced training - improved direction accuracy 55% -> 65%, down prediction 40% -> 50%
xianqiu
-
Add einops dependency
-
Deploy Kronos BTC Prediction API
- FastAPI server with /predict and /signal endpoints
- Kronos model (iter5, 62.4% win rate, 3.25 Sharpe ratio)
- Python client SDK
- Docker configuration for HuggingFace Spaces -
initial commit