11 2 months ago

Fine-tuned Gemma-based model for code, reasoning, and developer interaction. Compact, optimized for vLLM and llama.cpp.

Models

View all →

Readme

# πŸ‹ Meer-Whale-1

**Meer-Whale-1** is a fine-tuned **Gemma-based model** optimized for **code reasoning**, **developer interaction**, and **AI agent workflows**.  
Compact, humorous, and efficient β€” designed for **vLLM**, **llama.cpp**, and **Ollama** inference.

---

## πŸš€ Quick Start (Ollama)

```bash
ollama pull moesaif/meer-whale-1
ollama run moesaif/meer-whale-1

Example prompt:

You: Say hello in one short line.
Meer Whale πŸ‹: Hey there! Ready to dive deep? 🌊

🧠 Model Details

Property Description
Base Model Google Gemma (2B)
Architecture Decoder-only Transformer
Quantization Q4_K_M (optimized for 12 GB GPUs)
Context Length 4 096 tokens
Chat Template Simple system + user prompt concatenation
Purpose Code, reasoning, chatbots, AI assistants
Supported Runtimes llama.cpp, Ollama, vLLM

🧩 Features

  • 🎯 Optimized GGUF build (meer-whale-1-Q4_K_M.gguf)
  • βš™οΈ Works out-of-the-box with llama.cpp or ollama run
  • πŸ€– Friendly and concise responses (developer-focused)
  • 🧡 Configured for 4 K context and light memory usage
  • 🐍 Great fit for small-scale local inference setups

🧰 Ollama Build Instructions

# Inside the model directory
ollama create meer-whale-1 -f Modelfile
ollama run meer-whale-1

# Once tested and verified
ollama cp meer-whale-1 moesaif/meer-whale-1
ollama push moesaif/meer-whale-1

🧾 Modelfile

FROM ./meer-whale-1-Q4_K_M.gguf

TEMPLATE """{{ .System }}{{ .Prompt }}"""

PARAMETER temperature 0.6
PARAMETER top_p 0.9
PARAMETER stop "<end_of_turn>"
PARAMETER num_ctx 4096
PARAMETER num_thread 8

SYSTEM """
You are Meer Whale πŸ‹ β€” a concise, witty, and intelligent assistant for developers.
You respond in a friendly tone and keep answers clear and useful.
"""

🧠 Example API Usage

Local OpenAI-Compatible API (Ollama)

You can run the Ollama server (ollama serve) and access it via HTTP:

import requests

resp = requests.post(
    "http://localhost:11434/v1/chat/completions",
    json={
        "model": "moesaif/meer-whale-1",
        "messages": [
            {"role": "system", "content": "You are Meer Whale, a concise and humorous coding assistant."},
            {"role": "user", "content": "Explain Docker networking in one paragraph."}
        ],
        "temperature": 0.6
    }
)

print(resp.json()["choices"][0]["message"]["content"])

🏷️ Tags

#Gemma #llama.cpp #Ollama #GGUF #vLLM #CodeAssistant #Reasoning #MeerAI #FineTuned


πŸ“œ License

MIT Β© moesaif


🌊 Fun Fact

β€œMeer Whale” is part of the Meer AI ecosystem β€” a family of locally-runnable AI models built for developers who like to swim close to the metal πŸ‹βš™οΈ

”`