fineweb2-hq-tokenizers-v2
Collection
182 items β’ Updated
A Byte-Level BPE tokenizer trained on rus_Cyrl data from Fineweb-2-HQ.
| Parameter | Value |
|---|---|
| Algorithm | Byte-Level BPE |
| Language | rus_Cyrl |
| Target Vocab Size | 8,000 |
| Final Vocab Size | 8,000 |
| Pre-tokenizer | custom:rus_Cyrl |
| Number handling | individual |
| Contraction handling | True |
| Normalizer | NFC |
| Special Tokens | <s>, </s>, <pad>, <unk> |
| Training Shards | 2 |
from transformers import AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained("flexitok/bpe_rus_Cyrl_8000_v2")
tokens = tokenizer.encode("Hello, world!")
tokenizer.json β Full HuggingFace tokenizervocab.json β Vocabulary mappingmerges.txt β BPE merge rules| Text | Tokens | Token IDs |
|---|---|---|
Hello, world! 12345 This is a test. γγγ«γ‘γ― |
H, ell, o, ,, Δ , w, or, ld, !, Δ , 1, 2, 3, 4, 5, Δ , Th, is, Δ , is |
43, 4190, 82, 15, 178, 90, 815, 5338, 4, 178, 20, 21, 22, 23, 24, 178, 5456, 1062, 178, 1062 |