|
|
---
|
|
|
license: other
|
|
|
language:
|
|
|
- en
|
|
|
base_model:
|
|
|
- meta-llama/Llama-3.1-8B
|
|
|
new_version: shiv1119/ollama-model-shiv
|
|
|
pipeline_tag: text-generation
|
|
|
tags:
|
|
|
- ollama
|
|
|
- local-llm
|
|
|
- docker
|
|
|
- offline
|
|
|
- shiv
|
|
|
---
|
|
|
# 🦙 Shiv's Custom Local LLM (Ollama Ready) |
|
|
|
|
|
Welcome to **`ollama-model-shiv`**, a custom-built local language model designed to run **completely offline** using [Ollama](https://ollama.com/). This repo packages everything — from `Modelfile` to model blobs — ready for Docker-based deployment and local inference. |
|
|
|
|
|
--- |
|
|
## Link to model - https://huggingface.co/shiv1119/ollama-model-shiv |
|
|
|
|
|
## 📁 Directory Structure |
|
|
|
|
|
OllamaModelBuild/ |
|
|
├── model_test.py # Sample Python script to interact with the model |
|
|
├── .gitattributes # Git LFS or text handling config |
|
|
├── ollama_clean/ |
|
|
│ ├── docker-compose.yml # Docker config to run Ollama with this model |
|
|
│ ├── Modelfile # Ollama build instructions |
|
|
│ └── models/ # Model weights & metadata |
|
|
│ ├── blobs/ # Binary blobs of the model |
|
|
│ └── manifests/ # Manifest describing model structure |
|
|
|
|
|
|
|
|
--- |
|
|
|
|
|
## 🚀 Features |
|
|
|
|
|
- ✅ 100% Offline — No internet or API key needed |
|
|
- 🐋 Docker-ready with `docker-compose.yml` |
|
|
- ⚡ Works with [Ollama CLI](https://ollama.com) |
|
|
- 🔁 Full model reproducibility via blobs and manifests |
|
|
- 🧠 Based on `LLaMA2` / open-weight LLM architecture |
|
|
|
|
|
--- |
|
|
|
|
|
## 🛠️ Getting Started |
|
|
|
|
|
### 🔧 1. Install Prerequisites |
|
|
|
|
|
- [Install Docker](https://docs.docker.com/get-docker/) |
|
|
- (Optional) [Install Ollama CLI](https://ollama.com/download) — used during build |
|
|
|
|
|
### 🐋 2. Run the Model Using Docker |
|
|
|
|
|
In the root of the project: |
|
|
|
|
|
```bash |
|
|
cd ollama_clean |
|
|
docker-compose up --build |
|
|
``` |
|
|
|
|
|
This builds and runs the model container locally with your custom blobs and Modelfile. |
|
|
🧪 Test the Model (Optional) |
|
|
```bash |
|
|
docker-compose exec ollama ollama run tinyllama "Hello" |
|
|
``` |
|
|
Use the included Python script to test interaction: |
|
|
|
|
|
python model_test.py |
|
|
|
|
|
Customize it to query your local Ollama model running at http://localhost:11434. |
|
|
🧰 Model Components |
|
|
|
|
|
Modelfile: Blueprint for Ollama to build the model |
|
|
|
|
|
blobs/: Raw model weights |
|
|
|
|
|
manifests/: Metadata describing model format/version |
|
|
|
|
|
docker-compose.yml: Encapsulates build/run config |
|
|
|
|
|
🧠 About Ollama |
|
|
|
|
|
Ollama makes it simple to run LLMs locally on your own machine — private, fast, and API-free. |
|
|
|
|
|
📦 Repo Purpose |
|
|
|
|
|
This repository was created to: |
|
|
|
|
|
Host a working local LLM solution |
|
|
|
|
|
Enable offline inference |
|
|
|
|
|
Serve as a template for packaging custom models with Ollama |
|
|
|
|
|
📜 License |
|
|
|
|
|
This repo is for educational/research purposes only. Please ensure you comply with the license of any base models used (e.g., LLaMA2, Mistral, etc.). |
|
|
🙌 Credits |
|
|
|
|
|
Crafted with ❤️ by Shiv Nandan Verma |
