No AI Compute Devices
Your production server does not have a compatible GPU, or no external compute device has been registered.
Click + Register This Device, then run the printed commands from apps/sdrra-ai-worker (system Python at the repo root has no sdrra_worker module). Browser sessions never become training nodes.
Compute Selection Mode
Local FastAPI monitor is idle — use Training Compute above for remote GPU jobs
- Production Render AI has no CUDA GPU. Dispatch training to a registered compute agent.
- Closing the browser does not stop a remote job — the agent keeps heartbeating to the API.
- After training, deploy
best.ptto Render viaYOLO_WEIGHTS_URLor the model registry.
Epoch
— / 100
ETA
—
Idle
GPU Usage
—
No CUDA device
VRAM
—
Speed
—
Est. from dataset Ă— epochs
Best mAP@50
—
—
Current Loss
—
Training Status
Idle
Waiting…
Dataset Overview
| Training Images | — |
| Validation Images | — |
| Test Images | — |
| Classes | — |
| Annotations (est.) | — |
| Avg. / Image | — |
Class Distribution
Loading chart…
- Class counts appear after the AI training config loads a dataset.
Training Metrics (Live)
Box Loss
Waiting…
Cls Loss
Waiting…
DFL Loss
Waiting…
mAP@50
Waiting…
Precision
Waiting…
Recall
Waiting…
F1 Score
Waiting…
Learning Rate
Waiting…
Loading metrics…
Training Console
IdleNo console output yet.Output appears here once you run
python train.pySystem Monitor
GPU—
VRAM—
CPU—
RAM—
Notifications
- Start a run to see events
Data Augmentation Preview
Original
Rotation
Brightness
Blur
Mosaic
MixUp
HSV
Prediction Preview
Ground Truth
Pending
Prediction
Pending
Overlay
Pending
Heatmap
Pending
Confidence
- mAP@50—
- Precision—
- Recall—
- F1—
Confusion Matrix
Generates after validation
Precision-Recall Curve
Generates after validation
Best Checkpoints
| Epoch | mAP | P / R | Act |
|---|---|---|---|
| No checkpoint yet | |||
Model Comparison
| Model | mAP | P | R |
|---|---|---|---|
| Current run | — | — | — |