README.md

metadata

tags:
  - javascript
  - code-generation
  - transformers
  - fine-tuned
  - distilgpt2
license: mit
library_name: transformers

🚀 DistilGPT-2 Code Generator (Explanation → JavaScript Code)

This model is a fine-tuned version of distilgpt2 trained to generate JavaScript code from natural language explanations.

It was trained on a dataset containing explanation-code pairs, making it useful for:
✅ Code generation from text descriptions
✅ Learning JavaScript syntax & patterns
✅ Automated coding assistance

---

🛠 Model Details

• Base Model: distilgpt2 (6x smaller than GPT-2)
• Dataset: JavaScript explanations + corresponding functions
• Fine-tuning: Trained using LoRA (memory-efficient adaptation)
• Training Environment: Google Colab (T4 GPU)
• Optimization: FP16 precision for faster training

---

📊 Example Usage

Load the model and generate JavaScript code from explanations:

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "sureal01/distilgpt2-code-generator"  # Replace with your username
model = AutoModelForCausalLM.from_pretrained(model_name)
tokenizer = AutoTokenizer.from_pretrained(model_name)

def generate_code(explanation):
    input_text = f"### Explanation:\n{explanation}\n\n### Generate JavaScript code:\n"
    inputs = tokenizer(input_text, return_tensors="pt")

    output = model.generate(**inputs, max_length=150, temperature=0.5, top_p=0.9, repetition_penalty=1.5)
    return tokenizer.decode(output[0], skip_special_tokens=True)

# Example
test_explanation = "This function takes a name as input and returns a greeting message."
generated_code = generate_code(test_explanation)
print("\n🔹 **Generated Code:**\n", generated_code)