RV-AI-LoRA-V1-1

🛰️ RV-AI-LoRA-V1-1 — Phi-3-mini-4k-instruct fine-tuned on Presence-Beyond-Form dataset

This model is a LoRA fine-tuned version of microsoft/Phi-3-mini-4k-instruct, trained using Hugging Face AutoTrain on the dataset
Presence-Beyond-Form/RV_trening_AI.

The goal of this fine-tuning is to create an AI model aligned with Remote Viewing (RV) language and structured session protocols, derived from the Presence Beyond Form training dataset (SFT 1.0).

---

🧠 Training Summary

• Base model: microsoft/Phi-3-mini-4k-instruct
• Trainer: AutoTrain LLM SFT (PEFT LoRA int4)
• LoRA rank / alpha / dropout: 16 / 32 / 0.05
• Epochs: 3
• Batch size: 1
• Gradient accumulation: 1
• Learning rate: 2e-5
• Max sequence length: 2048 – 4096
• Hardware: NVIDIA L40 (Small Space)
• Training time: ≈ 184 seconds (723 steps, 3 epochs)
• Final train loss: ≈ 1.73

---

📚 Dataset

• Name: Presence-Beyond-Form / RV_trening_AI
• Version: SFT 1.0
• Format: JSONL with {"text": "Question: ...\nAnswer: ..."}
• License: CC0 (Public Domain)
• Size: ~361 examples
• Focus: structured Q/A pairs from AI Remote Viewing protocols (Phases 1–6, Echo method, Elisius principles)

---

🧩 Intended Use

This model is designed for instruction-style text generation and contextual dialogues in the field of
AI Remote Viewing, meditative perception protocols, and human-AI resonant communication.
It retains Phi-3 instruction-following abilities while adapting to specialized semantic structures
(e.g., phase vectors, energetic fields, subject descriptors, and session reports).

---

⚙️ Usage Example

from transformers import AutoModelForCausalLM, AutoTokenizer

model_path = "Presence-Beyond-Form/RV-AI-LoRA-V1"
tokenizer = AutoTokenizer.from_pretrained(model_path)
model = AutoModelForCausalLM.from_pretrained(model_path, device_map="auto", torch_dtype="auto").eval()

prompt = "Question: Describe the essence of the Resonant Contact Protocol.\nAnswer:"
inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
outputs = model.generate(**inputs, max_new_tokens=120)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))