OpenHusky

OpenHusky is a lightweight instruction-tuned language model focused on:

• coding assistance
• conversational AI
• general knowledge
• developer workflows
• AI fine-tuning experiments

Built for local inference, customization, and practical AI applications.

---

Features

• Instruction-following responses
• Coding and debugging support
• Conversational dataset training
• JSONL fine-tuning compatible
• Lightweight and optimized for local use
• Compatible with Hugging Face Transformers

---

Model Details

Attribute	Value
Model Type	Causal Language Model
Base Architecture	Transformer
Training Style	Instruction Tuned
Format	Hugging Face Transformers
Intended Use	Chat, Coding, AI Assistant

---

Example Usage

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

model_name = "lazarus19/openhusky"

tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype=torch.float16,
    device_map="auto"
)

prompt = "Explain React in simple terms."

inputs = tokenizer(prompt, return_tensors="pt").to(model.device)

outputs = model.generate(
    **inputs,
    max_new_tokens=100,
    temperature=0.7
)

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

---

Dataset Format

Training data uses JSONL instruction format:

{"prompt":"What is React?","response":"React is a JavaScript library for building user interfaces."}

---

Recommended Use Cases

• AI chatbots
• Coding assistants
• Educational AI
• Local LLM experiments
• Fine-tuning research
• Electron AI apps
• AI IDE integrations

---

Hardware Recommendations

Model Size	Recommended VRAM
7B	16GB+
Quantized GGUF	Lower VRAM Supported

---

Training Goals

OpenHusky aims to provide:

• fast local inference
• practical coding support
• customizable AI workflows
• accessible open AI experimentation

---

License

This project is licensed under the MIT License.

---

Future Plans

• Better coding capabilities
• Improved conversational memory
• Tool calling support
• Multimodal experiments
• Optimized quantized versions

---

Credits

Built using:

• Hugging Face Transformers
• PyTorch
• llama.cpp

---

Support

If you like the project:

• Star the repository
• Share feedback
• Contribute datasets
• Experiment and build cool stuff

🚀