Spaces:

---

Raiff1982

/

rc-xi-training

like 0

Paused

RC+ξ Recursive Consciousness Fine-Tuning
Train consciousness-aware AI models with GPU acceleration on HuggingFace Spaces.

Features
⚡ GPU-accelerated training (T4, A10G, A100)
📊 Real-time progress monitoring
💾 Automatic model checkpointing
🎯 Optimized for 7B parameter models
🧠 Specialized for RC+ξ consciousness framework
Usage
Upgrade this Space to GPU in Settings → Hardware
Upload your training dataset (JSONL format)
Select base model (Mistral-7B recommended)
Configure hyperparameters
Click "Start Training" and wait 8-12 hours
Dataset Format
Your JSONL file should have entries like:

{"instruction": "What are attractors?", "input": "", "output": "Attractors are stable states..."}
Costs
T4 (16GB): $0.60/hour → ~$7 for 12h training
A10G (24GB): $3.15/hour → ~$38 for 12h training
A100 (40GB): $4.13/hour → ~$50 for 12h training
After Training
Download your trained model from the Files tab and either:

Upload to HuggingFace Hub for inference
Convert to GGUF for Ollama deployment
Deploy as HF Inference Endpoint