Dashboard
System overview and quick access
Quick Actions
Active Modules
Script Builder
Drag and drop blocks to create HyperNix workflows
Blocks
Configuration
Select a block to configure
Training
Configure and monitor model training
Configuration
Telemetry
Quantize
GGUF quantization studio
Tupperware
Dataset round splitting with per-round step budgets
Tvtop Monitor
Real-time process and resource monitoring
Active Processes
| PID | User | GPU% | RAM% | Command | Action |
|---|---|---|---|---|---|
| Loading processes... | |||||
GPU Usage
System Load
Network Manager
Tailscale and network configuration
--
Network Interfaces
Modules
Available HyperNix components
All Modules
Module Reference
PressureCookerV3
Advanced optimizer with quantization-aware training (QAT), gradient checkpointing, adaptive clipping, EMA shadowing, and distributed awareness.
StovetopV3CookerPlus
Enhanced stovetop variant optimized for stable convergence on consumer hardware.
Abbicus
Dynamic curriculum learning system that automatically adjusts sequence length and padding based on model size and dataset complexity.
ComputeFramework
Hardware abstraction layer supporting CUDA, MPS, CPU, and TPU backends with automatic DDP/ZeRO wrapping.
Tupperware
Dataset round-splitting utility for multi-phase training with per-round step budgets and learning rate schedules.
HyperNixQuantizer
GGUF quantization pipeline supporting 30+ quantization types from F32 to IQ2_XXS.
Assistant Chat
AI-powered help for HyperNix workflows
Settings
Server configuration and preferences