README.MD

---
license: apache-2.0
base_model: Qwen/Qwen3-Coder-1.5B
tags:
- causal-lm
- qwen
- qwen3
- code
- coder
- lora-merged
- code-analysis
pipeline_tag: text-generation
---

# Code_analyze_1.0

**Code_analyze_1.0** is a merged LoRA fine-tuned version of **Qwen3-Coder-1.5B**, optimized for
code analysis, code understanding, and reasoning over source code.

## Model Details

- **Base model:** Qwen/Qwen3-Coder-1.5B  
- **Model type:** Causal Language Model  
- **Fine-tuning method:** LoRA (merged into base weights)  
- **Languages:** Primarily English (code-focused), supports multilingual comments  
- **Domain:** Programming / Software Engineering  

This model is **fully merged and standalone** — no additional LoRA adapters or base model
dependencies are required at inference time.

## Intended Use

The model is designed for:
- Static code analysis
- Bug detection and explanation
- Code review and refactoring suggestions
- Understanding unfamiliar codebases
- Explaining algorithms and logic in source code

## Usage

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "Vilyam888/Code_analyze_1.0"

tokenizer = AutoTokenizer.from_pretrained(
    model_id,
    trust_remote_code=True
)

model = AutoModelForCausalLM.from_pretrained(
    model_id,
    trust_remote_code=True,
    device_map="auto"
)

prompt = "Analyze this Python function and find potential issues:\n\n```python\ndef f(x): return x + 1\n```"
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=256)

print(tokenizer.decode(outputs[0], skip_special_tokens=True))