Upload README (2).md
Browse files- README (2).md +106 -0
README (2).md
ADDED
|
@@ -0,0 +1,106 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
language:
|
| 4 |
+
- ru
|
| 5 |
+
- en
|
| 6 |
+
- uk
|
| 7 |
+
---
|
| 8 |
+
### **🔥 PyroNet-mini**
|
| 9 |
+
|
| 10 |
+
**PyroNet-mini** is a fine-tuned lightweight variant of [microsoft/Phi-4-mini-instruct](https://huggingface.co/microsoft/Phi-4-mini-instruct),
|
| 11 |
+
trained with a **custom multilingual instruction dataset** (English, Ukrainian, Russian).
|
| 12 |
+
It belongs to the **PyroNet family**, serving as the **mini version** of the main [PyroNet](https://huggingface.co/Kenan023214/PyroNet).
|
| 13 |
+
|
| 14 |
+
---
|
| 15 |
+
|
| 16 |
+
### ✨ Features
|
| 17 |
+
- 🔹 Based on **Phi-4-mini-instruct**
|
| 18 |
+
- 🔹 Fine-tuned with **LoRA adapters**
|
| 19 |
+
- 🔹 Multilingual (English / Українська / Русский)
|
| 20 |
+
- 🔹 Softer safety alignment for more natural answers
|
| 21 |
+
- 🔹 Optimized for dialogue, coding help, and educational purposes
|
| 22 |
+
|
| 23 |
+
---
|
| 24 |
+
|
| 25 |
+
### 🚀 Usage
|
| 26 |
+
|
| 27 |
+
## Inference with Transformers
|
| 28 |
+
```python
|
| 29 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
| 30 |
+
import torch
|
| 31 |
+
|
| 32 |
+
model_id = "Kenan023214/PyroNet-mini"
|
| 33 |
+
|
| 34 |
+
tokenizer = AutoTokenizer.from_pretrained(model_id)
|
| 35 |
+
model = AutoModelForCausalLM.from_pretrained(
|
| 36 |
+
model_id,
|
| 37 |
+
device_map="auto",
|
| 38 |
+
torch_dtype="auto"
|
| 39 |
+
)
|
| 40 |
+
|
| 41 |
+
messages = [
|
| 42 |
+
{"role": "system", "content": system_prompt},
|
| 43 |
+
{"role": "user", "content": "Explain quantum entanglement in simple words."}
|
| 44 |
+
]
|
| 45 |
+
|
| 46 |
+
|
| 47 |
+
input_text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
|
| 48 |
+
inputs = tokenizer(input_text, return_tensors="pt").to(model.device)
|
| 49 |
+
|
| 50 |
+
# Generate the response
|
| 51 |
+
outputs = model.generate(
|
| 52 |
+
inputs,
|
| 53 |
+
max_new_tokens=256,
|
| 54 |
+
pad_token_id=tokenizer.eos_token_id
|
| 55 |
+
)
|
| 56 |
+
|
| 57 |
+
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
|
| 58 |
+
```
|
| 59 |
+
|
| 60 |
+
---
|
| 61 |
+
|
| 62 |
+
|
| 63 |
+
### 📊 Intended Use
|
| 64 |
+
|
| 65 |
+
Educational Q&A
|
| 66 |
+
|
| 67 |
+
Coding assistance
|
| 68 |
+
|
| 69 |
+
Multilingual chatbots
|
| 70 |
+
|
| 71 |
+
Creative writing and storytelling
|
| 72 |
+
|
| 73 |
+
|
| 74 |
+
|
| 75 |
+
---
|
| 76 |
+
|
| 77 |
+
### ⚡ Limitations
|
| 78 |
+
|
| 79 |
+
Being a mini model, it is less capable than larger PyroNet variants
|
| 80 |
+
|
| 81 |
+
May produce factual inaccuracies
|
| 82 |
+
|
| 83 |
+
Softer censorship may allow sensitive content
|
| 84 |
+
|
| 85 |
+
|
| 86 |
+
|
| 87 |
+
---
|
| 88 |
+
|
| 89 |
+
👤 Author & Contact
|
| 90 |
+
|
| 91 |
+
Author: IceL1ghtning
|
| 92 |
+
|
| 93 |
+
Maintainer: KenanAI
|
| 94 |
+
|
| 95 |
+
Contact:
|
| 96 |
+
|
| 97 |
+
|
| 98 |
+
<p align="left">
|
| 99 |
+
<a href="https://t.me/LogovoOfEngineer">
|
| 100 |
+
<img src="https://img.shields.io/badge/Telegram-Contact-blue?logo=telegram&style=for-the-badge" />
|
| 101 |
+
</a>
|
| 102 |
+
<a href="mailto:engineerglab@gmail.com">
|
| 103 |
+
<img src="https://img.shields.io/badge/Email-Contact-red?logo=gmail&style=for-the-badge" />
|
| 104 |
+
</a>
|
| 105 |
+
</p>
|
| 106 |
+
---
|