|
|
--- |
|
|
license: mit |
|
|
datasets: |
|
|
- GainEnergy/gpt-4o-oilandgas-trainingset |
|
|
base_model: |
|
|
- qihoo360/TinyR1-32B-Preview |
|
|
library_name: transformers |
|
|
tags: |
|
|
- oil-gas |
|
|
- drilling-engineering |
|
|
- retrieval-augmented-generation |
|
|
- finetuned |
|
|
- energy-ai |
|
|
- tiny-r1-32b |
|
|
- lora |
|
|
model-index: |
|
|
- name: OGAI-R1 |
|
|
results: |
|
|
- task: |
|
|
type: text-generation |
|
|
name: Oil & Gas Engineering AI |
|
|
dataset: |
|
|
name: GainEnergy GPT-4o Oil & Gas Training Set |
|
|
type: custom |
|
|
metrics: |
|
|
- name: Engineering Calculations Accuracy |
|
|
type: accuracy |
|
|
value: 94.3 |
|
|
- name: Technical Document Retrieval Precision |
|
|
type: precision |
|
|
value: 90.5 |
|
|
- name: Context Retention |
|
|
type: contextual-coherence |
|
|
value: High |
|
|
--- |
|
|
|
|
|
# OGAI-R1: Oil & Gas AI Model for Engineering & Technical Knowledge |
|
|
|
|
|
 |
|
|
[](LICENSE) |
|
|
|
|
|
**OGAI-R1** is a **fine-tuned version of TinyR1-32B**, designed specifically for **oil and gas engineering applications**. It is optimized for **engineering calculations, wellbore stability analysis, reservoir management, and document-based retrieval-augmented generation (RAG)**. |
|
|
|
|
|
The model has been trained using **GainEnergy's GPT-4o Oil & Gas Training Set**, incorporating expert knowledge, technical formulas, and structured query-response interactions. |
|
|
|
|
|
## 🏗 **Why Use OGAI-R1?** |
|
|
- **🚀 Fine-tuned for oil & gas engineering tasks** (drilling, production, reservoir, and refining). |
|
|
- **💡 Optimized for RAG** – Enhanced document understanding and retrieval. |
|
|
- **📚 Long-Context Retention** – Handles **up to 32K tokens** for complex engineering workflows. |
|
|
- **⚡ LoRA Fine-Tuning on TinyR1-32B** – Enables efficient inference and quick knowledge retrieval. |
|
|
|
|
|
--- |
|
|
|
|
|
## 🛠 **How to Use OGAI-R1** |
|
|
|
|
|
### **1️⃣ Install Required Dependencies** |
|
|
```bash |
|
|
pip install torch transformers accelerate bitsandbytes |
|
|
``` |
|
|
|
|
|
### **2️⃣ Load the Model** |
|
|
```python |
|
|
from transformers import AutoModelForCausalLM, AutoTokenizer |
|
|
|
|
|
model_id = "GainEnergy/OGAI-R1" |
|
|
|
|
|
# Load tokenizer |
|
|
tokenizer = AutoTokenizer.from_pretrained(model_id) |
|
|
|
|
|
# Load model |
|
|
model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto") |
|
|
|
|
|
# Run inference |
|
|
prompt = "Explain the principles of reservoir simulation in petroleum engineering." |
|
|
inputs = tokenizer(prompt, return_tensors="pt").to("cuda") |
|
|
outputs = model.generate(**inputs, max_new_tokens=512) |
|
|
print(tokenizer.decode(outputs[0], skip_special_tokens=True)) |
|
|
``` |
|
|
|
|
|
--- |
|
|
|
|
|
## 📦 **Model Variants** |
|
|
| **Model Name** | **Base Model** | **Precision** | **Context Window** | **Use Case** | |
|
|
|--------------|--------------|--------------|--------------|--------------| |
|
|
| **OGAI-R1** | TinyR1-32B | FP16 | 32K tokens | **Engineering Calculations & RAG** | |
|
|
| **OGAI-8x7B** | Mixtral-8x7B | 4-bit | 32K tokens | Oil & Gas AI Assistant | |
|
|
| **OGAI-Reasoner** | DeepSeek-R1 | FP16 | 128K tokens | Logical Reasoning & AI Simulation | |
|
|
|
|
|
--- |
|
|
|
|
|
## 📌 **Key Capabilities** |
|
|
✅ **Engineering Calculations** – Computes reservoir volumes, wellbore stability, mud weight, casing depth, and more. |
|
|
✅ **Technical Document Understanding** – Trained on oil and gas **technical literature, drilling reports, and engineering manuals**. |
|
|
✅ **Retrieval-Augmented Generation (RAG)** – Enhances AI-driven document retrieval for faster decision-making. |
|
|
✅ **High-Context Retention (32K tokens)** – Supports **long technical reports, operational workflows, and AI-driven engineering analysis**. |
|
|
|
|
|
--- |
|
|
|
|
|
## 🚀 **Use Cases** |
|
|
- **Wellbore Stability & Drilling Optimization** |
|
|
- **Hydraulics & Fluid Flow Simulations** |
|
|
- **Reservoir Engineering & Petrophysics Analysis** |
|
|
- **AI-Powered Document Retrieval & RAG Workflows** |
|
|
- **Technical Compliance & Regulatory Document Processing** |
|
|
|
|
|
--- |
|
|
|
|
|
## 📡 **Deployment Options** |
|
|
| **Platform** | **Compatible?** | **Recommended Setup** | |
|
|
|-------------|----------------|-----------------------| |
|
|
| **Hugging Face Inference API** | ✅ Yes | Deploy via `hf.co/GainEnergy/OGAI-R1` | |
|
|
| **RunPod.io (Serverless GPU)** | ✅ Yes | `A100-40GB` or `RTX 4090` | |
|
|
| **AWS EC2 (G5 Instances)** | ✅ Yes | `ml.g5.2xlarge` (8 vCPUs, 32GB RAM) | |
|
|
| **Local GPU (Consumer Hardware)** | ✅ Yes | Requires **≥16GB VRAM (RTX 3090, 4090)** | |
|
|
|
|
|
--- |
|
|
|
|
|
## ⚠️ **Limitations** |
|
|
🚧 **Optimized for Oil & Gas Engineering** – Not designed for general-purpose AI tasks. |
|
|
🚧 **Requires domain-specific expertise** – Outputs should be validated by industry experts. |
|
|
🚧 **Computational requirements** – Running the full TinyR1-32B model requires high-end GPUs. |
|
|
|
|
|
--- |
|
|
|
|
|
## 🔗 **Resources** |
|
|
- **[GainEnergy AI Platform](https://gain.energy)** – Explore AI-powered drilling automation. |
|
|
- **[Hugging Face Model Hub](https://huggingface.co/GainEnergy/OGAI-R1)** – Download & deploy the model. |
|
|
|
|
|
--- |
|
|
|
|
|
## 📚 **Citing OGAI-R1** |
|
|
```bibtex |
|
|
@article{ogai-r1-2025, |
|
|
title={OGAI-R1: An AI Model for Oil & Gas Engineering Optimization}, |
|
|
author={GainEnergy AI Team}, |
|
|
year={2025}, |
|
|
publisher={Hugging Face Models} |
|
|
} |
|
|
``` |
