File size: 2,833 Bytes
a3f0ccc
 
 
ac823ea
a3f0ccc
 
 
 
 
 
 
 
 
 
 
 
 
 
5bea2a1
a3f0ccc
 
 
5bea2a1
 
cb97d8a
a3f0ccc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
710b373
 
 
a3f0ccc
710b373
 
a3f0ccc
710b373
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
a3f0ccc
710b373
a3f0ccc
710b373
 
a3f0ccc
 
710b373
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
---
license: bsd-2-clause
datasets:
- hodza/BlackBox.Shkola.2014
language:
- ru
- en
base_model:
- Qwen/Qwen2.5-Coder-3B-Instruct
tags:
- code
- programming
- blackbox
- componentpascal
---

# BlackBox Component Pascal Assistant Model

![Model Logo](https://blackbox.oberon.org/blackbox.png) <!-- Optional logo -->

## Model Description

[Model repo on github](https://github.com/hodzanassredin/neuro-sft-host)

This is a specialized AI assistant for programming in **[BlackBox Component Builder](https://blackbox.oberon.org/download)** using Component Pascal. The model is fine-tuned on Qwen/Qwen2.5-Coder-3B-Instruct to provide context-aware coding assistance and best practices for BlackBox development.

**Key Features:**
- Component Pascal syntax support
- BlackBox framework-specific patterns
- Code generation and troubleshooting
- Interactive programming guidance

## Intended Use
✅ Intended for:
- BlackBox Component Builder developers
- Component Pascal learners
- Legacy Oberon-2 system maintainers
- Educational purposes

🚫 Not intended for:
- General programming outside BlackBox
- Non-technical decision making
- Mission-critical systems without human verification

## How to Use

```python
from transformers import BitsAndBytesConfig, AutoModelForCausalLM, AutoTokenizer
import torch
from peft import PeftModel

assert torch.cuda.is_available(), "you need cuda for this part"
device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')

base_model_name = 'Qwen/Qwen2.5-Coder-3B-Instruct'
qlora_adapter = "hodza/BlackBox-Coder-3B"
tokenizer = AutoTokenizer.from_pretrained(base_model_name)

bnb_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_quant_type="nf4",
    bnb_4bit_use_double_quant=True,
    bnb_4bit_compute_dtype=torch.bfloat16
)
base_model = AutoModelForCausalLM.from_pretrained(base_model_name, device_map=device,quantization_config=bnb_config,)

model = PeftModel.from_pretrained(base_model, qlora_adapter, device_map=device)
# Define the chat template
def format_chat_prompt(user_query):
    return [
        {"role": "system", "content": "You are a helpful coding assistant for BlackBox Component Builder using Component Pascal."},
        {"role": "user", "content": user_query}
    ]

def get_assistant_response(user_query):
    # Format the prompt using the chat template
    chat_prompt = format_chat_prompt(user_query)
    inputs = tokenizer.apply_chat_template(chat_prompt, return_tensors="pt").to(model.device)
    
    # Generate the response
    outputs = model.generate(
        inputs,
        max_new_tokens=256,
        temperature=0.3,
        top_p=0.3,
        pad_token_id=tokenizer.eos_token_id
    )
    return tokenizer.decode(outputs[0], skip_special_tokens=True)

print(get_assistant_response("Как мне вывести массив в Log?"))