Skip to content

fix(gguf): set ignore_merges for llama 3 gguf tokenizers #728

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
jason-fxz merged 1 commit into from
Oct 17, 2026
Merged

fix(gguf): set ignore_merges for llama 3 gguf tokenizers #728

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 5 additions & 0 deletions python/freetoken/models/gguf/tokenizer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22,6 +22,9 @@
"llama3", "llama-v3", "llama-bpe", "falcon3", "falcon-h1", "pixtral", "midm-2.0", "lfm2", "jina-v5-nano", "tekken", "chameleon",
))

# Llama 3 pre-tokenizers, whose HF tokenizer and llama.cpp take a word that is one vocabulary token as is instead of merging it
_IGNORE_MERGES_PRE = frozenset(("llama3", "llama-v3", "llama-bpe"))


def _add_bos(meta: dict) -> bool:
if "tokenizer.ggml.add_bos_token" in meta:
Expand Down Expand Up @@ -77,6 +80,8 @@ def load_gguf_tokenizer(model_path: str):
fast, _ = convert_gguf_tokenizer(arch, tok_dict)
meta = load_gguf_metadata(model_path)
fast = _fix_added_tokens(fast, meta)
if meta.get("tokenizer.ggml.model") == "gpt2" and meta.get("tokenizer.ggml.pre") in _IGNORE_MERGES_PRE:
fast.model.ignore_merges = True
# special: bos/eos/unk/pad as the file names them (None when absent) and the chat template
return PreTrainedTokenizerFast(
tokenizer_object=fast,
Expand Down
Loading