AxoLexis
🔬 Core Purpose:
AxoLexis is a fully-featured PyQt6 desktop trainer for the SHADA (Self-supervised Hierarchical Adaptive Hybrid Algorithm for Deep Learning) model.
🎨 UI Features:
• Real-time pyqtgraph charts — train loss, val loss, val accuracy, learning rate, gradient norm
• Dark theme stylesheet with splash screen and theme manager
• Multi-phase training control (pretrain, mtl, finetune, deploy, full_pipeline)
• Start/Stop/Save controls with background QThread execution
⚙️ Configuration:
• Select tier (nano → xl), task, optimizer, epochs, batch size, LR, LoRA rank, mixed precision
• Data Support: CSV, Image folders (ImageNet-style), NumPy .npy/.npz, built-in synthetic demo dataset
• Automatic NVIDIA GPU detection with user notification
• Optional PyInstaller .exe packaging for Windows
📁 Project Structure:
• main.py — Entry point with venv setup, splash screen, GPU check
• ui/ — Main window, data panel, model panel, training panel, plots panel, log panel
• training/ — QThread trainer worker
• models/ — Model factory
• data/ — Data loaders
Technical Stack
AxoLexis
PC Training Solution