instructions.md

Output Requirements

Important: Output the complete Python code directly in your reply:

1. Code must be wrapped in a python ... code block
2. Code must be runnable as-is, without modification
3. Do not include test code or benchmark code
4. Do not add extra explanations, output only the code block
5. Do not write code to a file, output it directly in your reply

Example output format:

import torch
import triton
import triton.language as tl

# Your implementation...