RealComp's picture
Upload 32 files
98657ab verified
|
Raw
History Blame Contribute Delete
1.78 kB
---
title: K1RL Quasar Volatility 50 (1s)
emoji: 💚
colorFrom: green
colorTo: gray
sdk: docker
pinned: false
---
# K1RL QUASAR — Quantitative Intelligence Observatory
> **HuggingFace Spaces Deployment** | 12GB RAM | Production-Ready AI Training Platform
## 🌌 Overview
K1RL QUASAR is a comprehensive AI training monitoring platform featuring:
- **Real-time Training Metrics** — Actor/Critic loss, AVN accuracy, buffer size
- **Advanced Health Monitoring** — RAM/disk management with auto-cleanup
- **HuggingFace Hub Integration** — Automatic checkpoint management
- **Sleep Prevention** — Keeps HF Spaces active 24/7
- **Multi-Service Architecture** — Redis, Flask, Supervisor orchestration
## 🚀 Services
- **Dashboard** → Port 7860 (HF Spaces standard)
- **Quasar Engine** → Main training loop
- **Health Monitor** → Resource management
- **Redis Server** → Real-time metrics
- **Sleep Prevention** → 24/7 uptime
## 📊 Dashboard Access
- **Main Interface**: `https://KarlQuant-k1rl-quasar-volatility-50-1s.hf.space/`
- **Health Check**: `https://KarlQuant-k1rl-quasar-volatility-50-1s.hf.space/health`
- **Metrics API**: `https://KarlQuant-k1rl-quasar-volatility-50-1s.hf.space/api/metrics`
## ⚙️ Required Secrets
Configure these in Space Settings → Secrets:
```
HF_TOKEN=your_token_here
HF_USERNAME=KarlQuant
REDIS_PASSWORD=k1rl_v501s_9d2f4a7c1e8b3065
```
## 🏗️ Architecture
Built with Docker + Supervisor managing multiple services:
- Flask dashboard on port 7860
- Redis for real-time metrics
- Automatic checkpoint backup to HF Hub
- 4-tier RAM monitoring (70% → 78% → 85% → 90%)
---
*K1RL QUASAR — Volatility 50 Index (1s) | Deriv | 8 Quantum AI Agents*