A fairly efficient *dupes implementation

pydupes.py

import collections
import os
import hashlib
import filecmp

def main():
    sizes = collections.defaultdict(list)

    for dirent in os.scandir():
        if not dirent.is_file(follow_symlinks=False):
            continue

        file_size = dirent.stat(follow_symlinks=False).st_size
        sizes[file_size].append(dirent.path)

    for same_sizes in sizes.values():
        if len(same_sizes) < 2:
            continue

        hashes = collections.defaultdict(list)

        for path in same_sizes:
            with open(path, "rb") as f:
                digest = hashlib.file_digest(f, hashlib.md5)
            file_hash = digest.digest()

            hashes[file_hash].append(path)

        for matches in hashes.values():
            while matches:
                printed_path0 = False
                path0 = matches.pop()
                for i in reversed(range(len(matches))):
                    path1 = matches[i]
                    if filecmp.cmp(path0, path1, shallow=False):
                        del matches[i]
                        if not printed_path0:
                            printed_path0 = True
                            print(path0, end='')
                        print(",", path1, end='')
                if printed_path0:
                    print()

if __name__ == "__main__":
    main()