|
|
--- |
|
|
license: mit |
|
|
language: |
|
|
- en |
|
|
- ru |
|
|
- ik |
|
|
--- |
|
|
# PyroNet-v1.5 |
|
|
|
|
|
**PyroNet-v1.5** is an advanced AI assistant built on top of **Microsoft Phi-2**. |
|
|
It introduces a custom **system persona prompt** through the tokenizer’s `chat_template`, enabling structured reasoning, multilingual support (Russian, English, Ukrainian), and a professional assistant style. |
|
|
|
|
|
## 🧠 Model Description |
|
|
- **Base model:** microsoft/phi-2 |
|
|
- **Languages:** Russian 🇷🇺, English 🇬🇧, Ukrainian 🇺🇦 |
|
|
- **Persona:** Professional, precise, calm senior engineer |
|
|
- **Reasoning:** Outputs step-by-step `Thought: ...` then `Final Answer: ...` |
|
|
|
|
|
## 🚀 Usage |
|
|
|
|
|
```python |
|
|
from transformers import AutoModelForCausalLM, AutoTokenizer |
|
|
import torch |
|
|
|
|
|
model_id = "Kenan023214/PyroNet-v1.5" |
|
|
|
|
|
tokenizer = AutoTokenizer.from_pretrained(model_id) |
|
|
model = AutoModelForCausalLM.from_pretrained( |
|
|
model_id, |
|
|
torch_dtype=torch.bfloat16, |
|
|
device_map="auto" |
|
|
) |
|
|
|
|
|
messages = [{"role": "user", "content": "Hello, who are you?"}] |
|
|
prompt = tokenizer.apply_chat_template(messages, add_generation_prompt=True, tokenize=False) |
|
|
inputs = tokenizer(prompt, return_tensors="pt").to(model.device) |
|
|
|
|
|
outputs = model.generate(**inputs, max_new_tokens=200, temperature=0.7, top_p=0.9) |
|
|
print(tokenizer.decode(outputs[0], skip_special_tokens=True)) |
|
|
``` |
|
|
## ⚠️ Limitations & Safety |
|
|
|
|
|
May produce hallucinations or inaccurate outputs. |
|
|
|
|
|
Not suitable for high-stakes domains (medical, legal, financial). |
|
|
|
|
|
Refuses unsafe or unethical instructions. |
