Skip to content

Add object detection, tracking, mobile notifications, and search to any security camera.

9.1 KB
Raw Blame
""" CLIP tokenizer

Copied from https://github.com/openai/CLIP. Originally MIT License, Copyright (c) 2021 OpenAI.
"""
#https://github.com/mlfoundations/open_clip/blob/d3cdb734a2710feeb4c6307df037afa5f786a3e1/src/open_clip/tokenizer.py
import gzip
import html
import os
import random
import string
from functools import lru_cache, partial
from typing import Callable, List, Optional, Union
import numpy as np
import unicodedata

# https://stackoverflow.com/q/62691279
os.environ["TOKENIZERS_PARALLELISM"] = "false"
_nltk_init = False

DEFAULT_CONTEXT_LENGTH = 77  # default context length for OpenAI CLIP

def is_letter(ch):
    return unicodedata.category(ch).startswith("L")

def is_number(ch):
    return unicodedata.category(ch).startswith("N")

def unicode_clip_tokenize(text, special_tokens):
    """
    Emulates the regex pattern from CLIP without using the 'regex' package:

        special | 's | 't | 're | 've | 'm | 'll | 'd | letters | numbers | other chars
    
    Returns a list of tokens exactly like CLIP regex would.
    """

    tokens = []
    i = 0
    n = len(text)
    special_sorted = sorted(special_tokens, key=len, reverse=True)

    while i < n:
        matched = False
        for sp in special_sorted:
            if text.startswith(sp, i):
                tokens.append(sp)
                i += len(sp)
                matched = True
                break
        if matched:
            continue

        ch = text[i]
        for cont in ["'s", "'t", "'re", "'ve", "'m", "'ll", "'d"]:
            if text.startswith(cont, i):
                tokens.append(cont)
                i += len(cont)
                matched = True
                break
        if matched:
            continue

        if is_letter(ch):
            start = i
            i += 1
            while i < n and is_letter(text[i]):
                i += 1
            tokens.append(text[start:i])
            continue

        if is_number(ch):
            tokens.append(ch)
            i += 1
            continue

        if not ch.isspace():
            start = i
            i += 1
            while (
                i < n
                and not text[i].isspace()
                and not is_letter(text[i])
                and not is_number(text[i])
            ):
                i += 1
            tokens.append(text[start:i])
            continue
        i += 1
    return tokens


@lru_cache()
def default_bpe():
    return os.path.join(os.path.dirname(os.path.abspath(__file__)), "bpe_simple_vocab_16e6.txt.gz")

@lru_cache()
def bytes_to_unicode():
    """
    Returns list of utf-8 byte and a corresponding list of unicode strings.
    The reversible bpe codes work on unicode strings.
    This means you need a large # of unicode characters in your vocab if you want to avoid UNKs.
    When you're at something like a 10B token dataset you end up needing around 5K for decent coverage.
    This is a significant percentage of your normal, say, 32K bpe vocab.
    To avoid that, we want lookup tables between utf-8 bytes and unicode strings.
    And avoids mapping to whitespace/control characters the bpe code barfs on.
    """
    bs = list(range(ord("!"), ord("~")+1))+list(range(ord("¡"), ord("¬")+1))+list(range(ord("®"), ord("ÿ")+1))
    cs = bs[:]
    n = 0
    for b in range(2**8):
        if b not in bs:
            bs.append(b)
            cs.append(2**8+n)
            n += 1
    cs = [chr(n) for n in cs]
    return dict(zip(bs, cs))


def get_pairs(word):
    """Return set of symbol pairs in a word.
    Word is represented as tuple of symbols (symbols being variable-length strings).
    """
    pairs = set()
    prev_char = word[0]
    for char in word[1:]:
        pairs.add((prev_char, char))
        prev_char = char
    return pairs


def basic_clean(text):
    text = html.unescape(html.unescape(text))
    return text.strip()


def whitespace_clean(text):
    text = " ".join(text.split())
    text = text.strip()
    return text


def _clean_canonicalize(x):
    # basic, remove whitespace, remove punctuation, lower case
    return canonicalize_text(basic_clean(x))


def _clean_lower(x):
    # basic, remove whitespace, lower case
    return whitespace_clean(basic_clean(x)).lower()


def _clean_whitespace(x):
    # basic, remove whitespace
    return whitespace_clean(basic_clean(x))


def get_clean_fn(type: str):
    if type == 'canonicalize':
        return _clean_canonicalize
    elif type == 'lower':
        return _clean_lower
    elif type == 'whitespace':
        return _clean_whitespace
    else:
        assert False, f"Invalid clean function ({type})."


def canonicalize_text(
    text,
    *,
    keep_punctuation_exact_string=None,
    trans_punctuation: dict = str.maketrans("", "", string.punctuation),
):
    """Returns canonicalized `text` (lowercase and punctuation removed).

    From: https://github.com/google-research/big_vision/blob/53f18caf27a9419231bbf08d3388b07671616d3d/big_vision/evaluators/proj/image_text/prompt_engineering.py#L94

    Args:
      text: string to be canonicalized.
      keep_punctuation_exact_string: If provided, then this exact string kept.
        For example providing '{}' will keep any occurrences of '{}' (but will
        still remove '{' and '}' that appear separately).
    """
    text = text.replace("_", " ")
    if keep_punctuation_exact_string:
        text = keep_punctuation_exact_string.join(
            part.translate(trans_punctuation)
            for part in text.split(keep_punctuation_exact_string)
        )
    else:
        text = text.translate(trans_punctuation)
    text = text.lower()
    text = " ".join(text.split())
    return text.strip()


class SimpleTokenizer(object):
    def __init__(
            self,
            bpe_path: str = default_bpe(),
            additional_special_tokens: Optional[List[str]] = None,
            context_length: Optional[int] = DEFAULT_CONTEXT_LENGTH,
            clean: str = 'lower',
            reduction_mask: str = ''
    ):
        self.byte_encoder = bytes_to_unicode()
        self.byte_decoder = {v: k for k, v in self.byte_encoder.items()}
        merges = gzip.open(bpe_path).read().decode("utf-8").split('\n')
        merges = merges[1:49152-256-2+1]
        merges = [tuple(merge.split()) for merge in merges]
        vocab = list(bytes_to_unicode().values())
        vocab = vocab + [v+'</w>' for v in vocab]
        for merge in merges:
            vocab.append(''.join(merge))
        special_tokens = ['<start_of_text>', '<end_of_text>']
        if additional_special_tokens:
            special_tokens += additional_special_tokens
        vocab.extend(special_tokens)
        self.encoder = dict(zip(vocab, range(len(vocab))))
        self.decoder = {v: k for k, v in self.encoder.items()}
        self.bpe_ranks = dict(zip(merges, range(len(merges))))
        self.cache = {t:t for t in special_tokens}
        special = "|".join(special_tokens)
        self.special_tokens_set = set(special_tokens)
        self.vocab_size = len(self.encoder)
        self.all_special_ids = [self.encoder[t] for t in special_tokens]
        self.sot_token_id = self.all_special_ids[0]
        self.eot_token_id = self.all_special_ids[1]
        self.context_length = context_length
        self.clean_fn = get_clean_fn(clean)
        self.reduction_fn = get_reduction_mask_fn(reduction_mask) if reduction_mask else None

    def bpe(self, token):
        if token in self.cache:
            return self.cache[token]
        word = tuple(token[:-1]) + ( token[-1] + '</w>',)
        pairs = get_pairs(word)

        if not pairs:
            return token+'</w>'

        while True:
            bigram = min(pairs, key = lambda pair: self.bpe_ranks.get(pair, float('inf')))
            if bigram not in self.bpe_ranks:
                break
            first, second = bigram
            new_word = []
            i = 0
            while i < len(word):
                try:
                    j = word.index(first, i)
                    new_word.extend(word[i:j])
                    i = j
                except Exception:
                    new_word.extend(word[i:])
                    break

                if word[i] == first and i < len(word)-1 and word[i+1] == second:
                    new_word.append(first+second)
                    i += 2
                else:
                    new_word.append(word[i])
                    i += 1
            new_word = tuple(new_word)
            word = new_word
            if len(word) == 1:
                break
            else:
                pairs = get_pairs(word)
        word = ' '.join(word)
        self.cache[token] = word
        return word

    def encode(self, text):
        bpe_tokens = []
        text = self.clean_fn(text)
        for token in unicode_clip_tokenize(text, self.special_tokens_set):
            token = ''.join(self.byte_encoder[b] for b in token.encode('utf-8'))
            bpe_tokens.extend(self.encoder[bpe_token] for bpe_token in self.bpe(token).split(' '))
        return bpe_tokens

    def decode(self, tokens):
        text = ''.join([self.decoder[token] for token in tokens])
        text = bytearray([self.byte_decoder[c] for c in text]).decode('utf-8', errors="replace").replace('</w>', ' ')
        return text


_tokenizer = SimpleTokenizer()