hodza/BlackBox-Coder-3B

BlackBox Component Pascal Assistant Model

Model Description

Model repo on github

This is a specialized AI assistant for programming in BlackBox Component Builder 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

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?"))