sevenhill/TradeBot
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Add Torch probe entries and Pi artifact sync

Browse Source
This commit is contained in:
Курнат Андрей
2026-06-24 21:31:05 +03:00
parent 15a50fb575
commit cb8efb7cd7
16 changed files with 307266 additions and 363669 deletions
Download Patch File Download Diff File Expand all files Collapse all files
README.md
+11 -1
View File
@@ -88,7 +88,13 @@ powershell -ExecutionPolicy Bypass -File tools\run_torch_retrain.ps1
powershell -ExecutionPolicy Bypass -File tools\install_windows_torch_retrainer.ps1
```
По умолчанию Windows-расписание переобучает PyTorch `LSTM/GRU` каждые 6 часов с `--limit 3000` на парах `BTCUSDT,ETHUSDT,SOLUSDT,LTCUSDT`. Параметры можно переопределить через env: `TORCH_RETRAIN_SYMBOLS`, `TORCH_RETRAIN_LIMIT`, `TORCH_RETRAIN_LOOKBACKS`, `TORCH_RETRAIN_ARCHITECTURES`, `TORCH_RETRAIN_HIDDEN_SIZES`, `TORCH_RETRAIN_LAYERS`, `TORCH_RETRAIN_DROPOUTS`, `TORCH_RETRAIN_HORIZON`, `TORCH_RETRAIN_HORIZONS`, `TORCH_RETRAIN_CONTEXT_SYMBOLS`, `TORCH_RETRAIN_FEATURES`, `TORCH_RETRAIN_EPOCHS`, `TORCH_RETRAIN_PATIENCE`, `TORCH_RETRAIN_INTERVAL`, `TORCH_RETRAIN_ENV`.
По умолчанию Windows-расписание переобучает PyTorch `LSTM/GRU` каждые 6 часов с `--limit 3000` на парах `BTCUSDT,ETHUSDT,SOLUSDT,LTCUSDT`. Параметры можно переопределить через env: `TORCH_RETRAIN_SYMBOLS`, `TORCH_RETRAIN_LIMIT`, `TORCH_RETRAIN_LOOKBACKS`, `TORCH_RETRAIN_ARCHITECTURES`, `TORCH_RETRAIN_HIDDEN_SIZES`, `TORCH_RETRAIN_LAYERS`, `TORCH_RETRAIN_DROPOUTS`, `TORCH_RETRAIN_HORIZON`, `TORCH_RETRAIN_HORIZONS`, `TORCH_RETRAIN_CONTEXT_SYMBOLS`, `TORCH_RETRAIN_FEATURES`, `TORCH_RETRAIN_SEED`, `TORCH_RETRAIN_EPOCHS`, `TORCH_RETRAIN_PATIENCE`, `TORCH_RETRAIN_INTERVAL`, `TORCH_RETRAIN_ENV`.
Если retrain запускается с `-DeployToPi`, после успешного guard он синхронизирует `runtime/lstm_forecaster.json`, `runtime/torch_retrain_guard.json` и `runtime/torch_threshold_calibration.json` на Raspberry Pi через SSH-ключ и перезапускает сервис `tradebot`. Отдельный запуск sync:
```powershell
powershell -ExecutionPolicy Bypass -File tools\sync_torch_artifacts_to_pi.ps1 -RemoteHost 192.168.0.185 -RemoteUser sevenhill -RemoteRoot /mnt/data/tradebot
```
Внутри recurrent модели используются exportable attention pooling и LayerNorm перед forecast-head; Raspberry Pi по-прежнему исполняет модель из JSON без PyTorch runtime.
@@ -165,6 +171,10 @@ TIME_SERIES_MIN_CONFIDENCE=0.4
TIME_SERIES_MAX_ADJUSTMENT=0.08
TIME_SERIES_LSTM_ENABLED=true
TIME_SERIES_LSTM_MODEL_PATH=runtime/lstm_forecaster.json
TIME_SERIES_PROBE_ENABLED=true
TIME_SERIES_PROBE_MIN_EDGE_PERCENT=0.02
TIME_SERIES_PROBE_MIN_PROBABILITY_UP=0.55
TIME_SERIES_PROBE_SIZE_MULTIPLIER=0.40
STOP_LOSS_PERCENT=0.04
TAKE_PROFIT_PERCENT=0.035
TRAILING_STOP_PERCENT=0.015