Datasets:

hf-internal-testing
/

tokenizers-test-data

Modalities:

Text

Formats:

text

Size:

< 1K

Libraries:

Datasets

Dataset card Data Studio Files Files and versions

xet

Community

Add SmolLM2-135M and bloomz-560m tokenizer.json fixtures

#14

by sbrandeis HF Staff - opened about 10 hours ago

base: refs/heads/main

←

from: refs/pr/14

Discussion Files changed

-0

sbrandeis

Hugging Face Internal Testing Organization org about 10 hours ago

Hugging Face CLI

Byte-identical tokenizer.json from HuggingFaceTB/SmolLM2-135M@93efa2f097d58c2a74874c7e644dbc9b0cee75a2 and bigscience/bloomz-560m@a2845d7e13dd12efae154a9f1c63fcc2e0cc4b05 (main as of 2026-11-06). Both are byte-level BPE vocabularies with no token for some bytes valid UTF-8 can contain (SmolLM2: 0x04 0x06 0x13 0x14 0x16 0x1D 0xF1 0xF2; bloomz: 0x0B). The tk-convert oracle in huggingface/tokenizers (PR #2550) checks they encode like the released crate, which drops those bytes.

Add SmolLM2-135M and bloomz-560m tokenizer.json fixtures40be28dd

sbrandeis changed pull request status to open about 10 hours ago

sbrandeis changed pull request status to merged about 10 hours ago

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment