Llama-3.2-1B-FastApply

How to use?

• We use Unsloth for faster inference and load the adapter:

from unsloth import FastLanguageModel
max_seq_length = 8192 
dtype = None 
load_in_4bit = True
model, tokenizer = FastLanguageModel.from_pretrained(
    model_name = "patched-codes/Llama-3.2-1B-FastApply",
    max_seq_length = max_seq_length,
    dtype = dtype,
    load_in_4bit = load_in_4bit,
)
FastLanguageModel.for_inference(model) # Enable native 2x faster inference

• The model works with original code and the edited code as input to generate the final updated code:

original_code = """import React from 'react';
import { Loader } from 'lucide-react';

interface ButtonProps {
  text: string;
  onClick?: () => void;
  loading?: boolean;
  disabled?: boolean;
  icon?: React.ReactNode;
}

const Button: React.FC<ButtonProps> = ({
  text,
  onClick,
  loading = false,
  disabled = false,
  icon
}) => (
  <button
    className="bg-blue-500 text-white p-2 rounded flex items-center gap-2"
    onClick={onClick}
    disabled={disabled || loading}
  >
    {loading ? <Loader className="animate-spin" /> : icon}
    {text}
  </button>
);

export default Button;
"""

update_snippet = """interface ButtonProps {
  variant?: 'primary' | 'secondary' | 'danger';
  size?: 'small' | 'medium' | 'large';
  // ... other props
}

const Button: React.FC<ButtonProps> = ({
  variant = 'primary',
  size = 'medium',
  // ... other props
}) => (
  <button
    className={`flex items-center gap-2 rounded ${
      size === 'small' ? 'p-1 text-sm' :
      size === 'large' ? 'p-3 text-lg' :
      'p-2 text-md'
    } ${
      variant === 'primary' ? 'bg-blue-500 text-white' :
      variant === 'secondary' ? 'bg-gray-500 text-white' :
      'bg-red-500 text-white'
    }`}
    // ... other attributes
  >
    // ... existing code ...
  </button>
);
"""

• Prepare your input following the prompt structure:

input_text = f"""
Merge all changes from the <update> snippet into the <code> below.
- Preserve the code's structure, order, comments, and indentation exactly.
- Output only the updated code, enclosed within <updated-code> and </updated-code> tags.
- Do not include any additional text, explanations, placeholders, ellipses, or code fences.

<code>{original_code}</code>

<update>{update_snippet}</update>

Provide the complete updated code.
"""

messages = [
    {"role": "system", "content": "You are a coding assistant that helps merge code updates, ensuring every modification is fully integrated."},
    {"role": "user", "content": input_text.strip()},
]

inputs = tokenizer.apply_chat_template(
    messages,
    tokenize = True,
    add_generation_prompt = True, # Must add for generation
    return_tensors = "pt",
).to("cuda")

from transformers import TextStreamer
text_streamer = TextStreamer(tokenizer, skip_prompt = True)
output = model.generate(input_ids = inputs, streamer = text_streamer, max_new_tokens = 8192,
                   use_cache = True, temperature = 1.5, min_p = 0.1)

response = tokenizer.decode(output[0][len(inputs[0]):])

updated_code = response.split("<updated-code>")[1].split("</updated-code>")[0]

Uploaded model

• Developed by: patched-codes
• License: apache-2.0
• Finetuned from model : unsloth/llama-3.2-1b-instruct-bnb-4bit

This llama model was trained 2x faster with Unsloth and Huggingface's TRL library.

<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>