How touhoufan2024/qlibassistant works
QlibAssistant automates CSI 300 market-data updates, rolling model training, ensemble prediction, and historical backtesting, then publishes prediction and review pages. The main user entry point is the roll CLI: it loads configuration and dispatches data, training, or model commands. Separate branches cover model archives, optional DoltHub CSV synchronization, and static-page generation. Edges are limited to relationships shown in excerpts or stated in the README; unsampled implementation details are not inferred.
Main entry points: modelcli.py Model selection and inference · roll.py Command dispatcher · model_review.py Backtest review · traincli.py Rolling training
Components 18
CLI and configuration
- Command dispatcher
roll/roll.py - Model configuration
roll/myconfig.py
Data and modeling
- Market data updates
roll/datacli.py - Local Qlib data
- Rolling training
roll/traincli.py - Model experiments
- Model selection and inference
roll/modelcli.py - Model archive management
roll/model_backup.py
Review and results
- Backtest review
roll/model_review.py - Prediction CSVs
qlib_score_csv - Review statistics
review_csv - Backtest results
backtest_csv
Publishing and sync
- Static page generation
page/script/gen_page.py - Results website
page/docs - CSV synchronization
script/sync_dolt.py
Other components
- CLI user
- Investment data releases
- DoltHub SQL API
Connections 23
- CLI user → Command dispatcher: runs commands
README.md - Command dispatcher → Market data updates: dispatches data
roll/roll.py - Command dispatcher → Rolling training: dispatches training
roll/roll.py - Command dispatcher → Model selection and inference: dispatches model
roll/roll.py - Market data updates → Investment data releases: downloads data
roll/datacli.py - Market data updates → Local Qlib data: updates data
README.md - Rolling training → Model configuration: builds tasks
roll/traincli.py - Rolling training → Local Qlib data: trains on data
README.md - Rolling training → Model experiments: records runs
roll/traincli.py - Model selection and inference → Model experiments: filters models
README.md - Model selection and inference → Local Qlib data: predicts from data
README.md - Model selection and inference → Prediction CSVs: writes scores
README.md - Command dispatcher → Model archive management: dispatches archive
- Model selection and inference → Model archive management: delegates archive
roll/modelcli.py - Command dispatcher → Backtest review: dispatches review
- Model selection and inference → Backtest review: delegates review
roll/modelcli.py - Backtest review → Model selection and inference: requests labels
roll/model_review.py - Backtest review → Prediction CSVs: reads predictions
roll/model_review.py - Backtest review → Review statistics: saves statistics
roll/model_review.py - Backtest review → Backtest results: saves backtests
roll/model_review.py - Static page generation → Backtest results: copies results
page/script/gen_page.py - Static page generation → Results website: generates pages
page/script/gen_page.py - CSV synchronization → DoltHub SQL API: syncs tables
script/sync_dolt.py