Cohere multilingual-22-12 tokenizer

This is the tokenizer for the Cohere multilingual-22-12 embedding model: Cohere Multilingual Embeddings

You can load it with the transformers library like this:

from transformers import AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained("CohereLabs/multilingual-22-12")
text = "Hellö World, this is my input string!"
enc = tokenizer(text)
print("Encoded input:")
print(enc)

inv_vocab = {v: k for k, v in tokenizer.vocab.items()}
tokens = [inv_vocab[token_id] for token_id in enc['input_ids']]
print("Tokens:")
print(tokens)

number_of_tokens = len(enc['input_ids'])
print("Number of tokens:", number_of_tokens)

Computing number of tokens

The following values can be used to approximate the number of tokens given the number input characters:

approx_number_of_tokens = len(input_text) / ratio

E.g. for English, approx_number_of_tokens = len(input_text) / 4.8.

| Language | Avg. characters per token |

| --- | :---: |

| ar | 3.6 |

| de | 4.6 |

| en | 4.8 |

| es | 4.6 |

| fr | 4.4 |

| hi | 3.8 |

| it | 4.5 |

| ja | 1.3 |

| ko | 2.0 |

| zh | 1.1 |

These values have been computed on the first 10,000 paragraphs from Wikipedia. For other dataset, these values might change.