license: apache-2.0
language:
- en
- ru
- uk
base_model:
- openai/gpt-oss-20b
🔥 PyroNet
PyroNet is a research-first open-source large language model, fine-tuned and customized with a unique system prompt and identity.
Created and maintained by IceL1ghtning from Ukraine 🇺🇦.
This model is based on gpt-oss-20b and inherits its architecture, but provides a different persona, behavior style, and chat template.
✨ Features
- 🧠 Custom system prompt defining the PyroNet identity
- 🎭 Optimized for conversational tasks (chat, reasoning, coding help)
- 🔗 Fully compatible with the Hugging Face
transformerslibrary - 📦 Distributed with a custom
chat_template.jinja
🚀 Usage
Install requirements
pip install transformers accelerate
Quick inference
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
model_id = "Kenan023214/PyroNet"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto")
pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)
prompt = "Hello, PyroNet! Can you introduce yourself?"
result = pipe(prompt, max_new_tokens=200, do_sample=True, temperature=0.8)
print(result[0]["generated_text"])
💡 Recommendations
Hardware: best run on a GPU with at least 24 GB VRAM (e.g. RTX 3090, A100).
For inference on smaller GPUs, use:
model = AutoModelForCausalLM.from_pretrained(
model_id,
device_map="auto",
load_in_8bit=True
)
(requires bitsandbytes).
Adjust temperature and top_p for creative or deterministic outputs.
⚡ System Prompt (Excerpt)
PyroNet is designed as a witty, research-driven assistant, capable of reasoning, coding, explaining decisions, and keeping conversations engaging.
📞 Contact
You can contact us via email: engineerglab@gmail.com
📜 License & Disclaimer
Based on gpt-oss-20b.
For research purposes only. Not intended for production or deployment without further alignment and safety checks.
Responsibility for usage lies with the end-user.
🔥 PyroNet — Where logic meets creativity.