app.py

from transformers import GPT2LMHeadModel, GPT2Tokenizer

# Load the pre-trained GPT2 model and tokenizer
model = GPT2LMHeadModel.from_pretrained("gpt2")
tokenizer = GPT2Tokenizer.from_pretrained("gpt2")

# 設置填充標記 ID
tokenizer.pad_token = tokenizer.eos_token

def generate_code(prompt, max_length=200):
    full_prompt = f"Generate Python code for {prompt}:```python\n"
    inputs = tokenizer.encode(full_prompt, return_tensors="pt")
    
    output = model.generate(
        inputs, 
        max_length=max_length, 
        num_return_sequences=1, 
        temperature=0.7,
        pad_token_id=tokenizer.pad_token_id
    )
    generated_text = tokenizer.decode(output[0], skip_special_tokens=False)
    
    # 提取生成的代碼
    start = generated_text.find("```python") + len("```python")
    end = generated_text.find("```", start)
    if end == -1:
        end = len(generated_text)
    code = generated_text[start:end].strip()
    
    return code

# Example usage
prompt = "a function to calculate the factorial of a number"
generated_code = generate_code(prompt)
print(generated_code)