Mungert/AQUA-7B-GGUF

AQUA-7B GGUF Models

Model Generation Details

This model was generated using llama.cpp at commit 21c02174.

---

Quantization Beyond the IMatrix

I've been experimenting with a new quantization approach that selectively elevates the precision of key layers beyond what the default IMatrix configuration provides.

In my testing, standard IMatrix quantization underperforms at lower bit depths, especially with Mixture of Experts (MoE) models. To address this, I'm using the --tensor-type option in llama.cpp to manually "bump" important layers to higher precision. You can see the implementation here:
👉 Layer bumping with llama.cpp

While this does increase model file size, it significantly improves precision for a given quantization level.

I'd love your feedback—have you tried this? How does it perform for you?

---

Click here to get info on choosing the right GGUF model format

---

Model Information

AQUA-7B is Kurma AI’s flagship 7-billion parameter large language model built exclusively for the global aquaculture industry. And it is the first large language model for the aquaculture. It is fine-tuned to deliver actionable insights for aquaculture species-specific farming, hatchery operations, water quality control, and disease management.

Trained on over 3 million real and synthetic aquaculture conversations (~1B tokens), AquaGPT-7B brings the power of domain-specific AI to fish farms, fish hatcheries, researchers, and Aqua-Tech innovators worldwide.

Learn more about Kurma AI.

---

Key Features

• Production Systems & Species Management: Covers ponds, tanks, cages, RAS, aquaponics, mariculture, and longlines. Delivers best practices for raising tilapia, catfish, carp, salmon, shrimp, crabs, oysters, trout, sea bass, and more supporting both smallholder and industrial farms.
• Genetics, Hatchery, and Early Life Stage Management: Guides advanced breeding, gene editing, hatchery design, spawning, larval care, nursery systems, live feed, transport, egg incubation, and biosecurity.
• Nutrition, Feeding, and Growth Optimization: Provides actionable protocols for water quality (temperature, oxygen, pH, ammonia, nitrite, salinity), and structured disease management: identification, vaccination, biosecurity, antibiotic use, and outbreak response.
• Water Quality, Health, and Disease Management: Provides actionable protocols for water quality (temperature, oxygen, pH, ammonia, nitrite, salinity), and structured disease management: identification, vaccination, biosecurity, antibiotic use, and outbreak response.
• Sustainable Aquaculture & Innovation: Promotes Promotes eco-friendly practices in waste management, environmental impact, biodiversity, climate adaptation, and guides adoption of new technologies AI, automation, sensors, water drones, and modern farm management.
• Water Quality, Health, and Disease Management: Advises on market trends, business planning, regulation, certification, traceability, and insurance. Covers best practices for harvesting, processing, cold chain, grading, packaging, contamination prevention, HACCP, and food safety.

---

Training Data Highlights

• Extension worker–farmer dialogues and field advisory logs
• FAO, ICAR, NOAA, and peer-reviewed aquaculture research
• Synthetic Q&A from 5,000+ aquaculture-focused topics
• Climate-resilient practices, hatchery SOPs, and water quality datasets
• Carefully curated to support species-specific culture methods
• Scale: Trained on approximately 3 million real and synthetic Q&A pairs, totaling around 1 billion tokens of high-quality, domain-specific data.

---

Model Specifications

• Base Model: Mistral 7B v0.3 (by Mistral AI)
• Training Tokens: ~1 Billion
• Released On 4, July 2025
• Data Volume: 3M+ expert-verified and synthetic instructions
• Origin: Made in America by Kurma AI
• Training Technic Model is trained via Fine-tuning using (LoRA-based) Supervised Fine-Tuning (SFT).
• Training Infrastructure: Trained using 16 NVIDIA H200 GPU Multi Cluster Special Thanks to Nebius

---

Quickstart

Transformers (Google Colab/ jupyter)

• Install dependencies

!pip install transformers accelerate

• Log in with your Hugging Face access token

from huggingface_hub import login

• Import model from Huggingface

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

model_id = "KurmaAI/AQUA-7B"

tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    device_map="auto",             # Automatically uses GPU if available
    torch_dtype=torch.float16,     # Use torch.float32 if no GPU
    trust_remote_code=True
)

• Test Prompt

prompt = "What are the most common diseases in shrimp farming and how can they be prevented?"
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=256)

response = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(response)

---

🙏 Acknowledgements

This project was made possible thanks to:

• Nebius for providing a compute grant and access to NVIDIA H200 GPU servers, which powered the model training process.
• Mistral for sharing their open-source language models, which made this project possible.
• Kurma AI research team: including aquaculture experts, machine learning engineers, data annotators, and advisors who collaborated to curate, verify, and refine the domain-specific dataset used for fine-tuning this model.

---

⚠️ Disclaimer, Bias & Limitations

• Domain Bias: The model may reflect inherent biases present in the aquaculture data sources and industry practices on which it was trained.
• Temporal Data Limitation: Climate and environmental recommendations are based on information available up to 2024. Users should cross-check any climate-related advice against the latest advisories (e.g., IMD or NOAA updates).
• Potential Hallucinations: Like all large language models, Aqua-7B may occasionally generate inaccurate or misleading responses ("hallucinations"). Always validate critical, regulatory, or high-impact decisions with a qualified aquaculture professional.

---

🚀 If you find these models useful

Help me test my AI-Powered Quantum Network Monitor Assistant with quantum-ready security checks:

👉 Quantum Network Monitor

The full Open Source Code for the Quantum Network Monitor Service available at my github repos ( repos with NetworkMonitor in the name) : Source Code Quantum Network Monitor. You will also find the code I use to quantize the models if you want to do it yourself GGUFModelBuilder

💬 How to test:
Choose an AI assistant type:

• TurboLLM (GPT-4.1-mini)
• HugLLM (Hugginface Open-source models)
• TestLLM (Experimental CPU-only)

What I’m Testing

I’m pushing the limits of small open-source models for AI network monitoring, specifically:

• Function calling against live network services
• How small can a model go while still handling:
  • Automated Nmap security scans
  • Quantum-readiness checks
  • Network Monitoring tasks

🟡 TestLLM – Current experimental model (llama.cpp on 2 CPU threads on huggingface docker space):

• ✅ Zero-configuration setup
• ⏳ 30s load time (slow inference but no API costs) . No token limited as the cost is low.
• 🔧 Help wanted! If you’re into edge-device AI, let’s collaborate!

Other Assistants

🟢 TurboLLM – Uses gpt-4.1-mini :

• **It performs very well but unfortunatly OpenAI charges per token. For this reason tokens usage is limited.
• Create custom cmd processors to run .net code on Quantum Network Monitor Agents
• Real-time network diagnostics and monitoring
• Security Audits
• Penetration testing (Nmap/Metasploit)

🔵 HugLLM – Latest Open-source models:

• 🌐 Runs on Hugging Face Inference API. Performs pretty well using the lastest models hosted on Novita.

💡 Example commands you could test:

1. "Give me info on my websites SSL certificate"
2. "Check if my server is using quantum safe encyption for communication"
3. "Run a comprehensive security audit on my server"
4. '"Create a cmd processor to .. (what ever you want)" Note you need to install a Quantum Network Monitor Agent to run the .net code on. This is a very flexible and powerful feature. Use with caution!

Final Word

I fund the servers used to create these model files, run the Quantum Network Monitor service, and pay for inference from Novita and OpenAI—all out of my own pocket. All the code behind the model creation and the Quantum Network Monitor project is open source. Feel free to use whatever you find helpful.

If you appreciate the work, please consider buying me a coffee ☕. Your support helps cover service costs and allows me to raise token limits for everyone.

I'm also open to job opportunities or sponsorship.

Thank you! 😊