Skip to main content
Training Playground
Interactive model training with real-time metrics and hyperparameter tuning
Hyperparameters Configuration
Model Architecture
CNN Small (1.2M params)
CNN + CBAM (2.5M params)
ResNet-18 (11M params)
ViT Tiny (5.7M params)
Hybrid CNN-ViT (8.3M params)
Dataset
Semiconductor Defects
Medical Imaging
Combined Dataset
Synthetic Data
Optimizer
Adam
AdamW
SGD with Momentum
RMSprop
Learning Rate Schedule
Constant
Step Decay
Cosine Annealing
One Cycle
Reduce on Plateau
Learning Rate
1e-3
Batch Size
32
Epochs
50
Dropout Rate
0.5
Weight Decay
1e-4
Mixed Precision (FP16)
Gradient Clipping
Data Augmentation
Input
→
Conv
→
CBAM
→
Pool
→
Conv
→
CBAM
→
FC
→
Output
Start Training
Stop
Save Model
Training Progress
Ready
0%
Live Metrics
0
Current Epoch
0.000
Training Loss
0.0%
0.0%
Train Accuracy
0.0%
0.0%
Val Accuracy
-
Learning Rate
1e-3
Time Elapsed
00:00
ETA
--:--
Training Log
System ready. Configure parameters and click Start Training.