Skip to content

Add pre-tokenizer regexes for BLOOM and gpt3-finnish #8850

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
merged 1 commit into from
Aug 15, 2024
Merged

Add pre-tokenizer regexes for BLOOM and gpt3-finnish #8850

Show file tree
Hide file tree
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
8 changes: 7 additions & 1 deletion convert_hf_to_gguf.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -603,6 +603,12 @@ def get_vocab_base_pre(self, tokenizer) -> str:
if chkhsh == "855059429035d75a914d1eda9f10a876752e281a054a7a3d421ef0533e5b6249":
# ref: https://huggingface.co/HuggingFaceTB/SmolLM-135M
res = "smollm"
if chkhsh == "3c30d3ad1d6b64202cd222813e7736c2db6e1bd6d67197090fc1211fbc612ae7":
# ref: https://huggingface.co/bigscience/bloom
res = "bloom"
if chkhsh == "bc01ce58980e1db43859146dc51b1758b3b88729b217a74792e9f8d43e479d21":
# ref: https://huggingface.co/TurkuNLP/gpt3-finnish-small
res = "gpt3-finnish"

if res is None:
logger.warning("\n")
Expand Down Expand Up @@ -906,7 +912,7 @@ def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iter
return tensors


@Model.register("BloomForCausalLM")
@Model.register("BloomForCausalLM", "BloomModel")
class BloomModel(Model):
model_arch = gguf.MODEL_ARCH.BLOOM

Expand Down
2 changes: 2 additions & 0 deletions convert_hf_to_gguf_update.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -94,6 +94,8 @@ class TOKENIZER_TYPE(IntEnum):
{"name": "codeshell", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/WisdomShell/CodeShell-7B", },
{"name": "tekken", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/mistralai/Mistral-Nemo-Base-2407", },
{"name": "smollm", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/HuggingFaceTB/SmolLM-135M", },
{'name': "bloom", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/bigscience/bloom", },
{'name': "gpt3-finnish", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/TurkuNLP/gpt3-finnish-small", },
]


Expand Down
2 changes: 2 additions & 0 deletions include/llama.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -93,6 +93,8 @@ extern "C" {
LLAMA_VOCAB_PRE_TYPE_TEKKEN = 20,
LLAMA_VOCAB_PRE_TYPE_SMOLLM = 21,
LLAMA_VOCAB_PRE_TYPE_CODESHELL = 22,
LLAMA_VOCAB_PRE_TYPE_BLOOM = 23,
LLAMA_VOCAB_PRE_TYPE_GPT3_FINNISH = 24,
};

// note: these values should be synchronized with ggml_rope
Expand Down
2 changes: 2 additions & 0 deletions src/llama-vocab.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -424,6 +424,8 @@ struct llm_tokenizer_bpe {
};
break;
case LLAMA_VOCAB_PRE_TYPE_PORO:
case LLAMA_VOCAB_PRE_TYPE_BLOOM:
case LLAMA_VOCAB_PRE_TYPE_GPT3_FINNISH:
regex_exprs = {
" ?[^(\\s|.,!?…。，、।۔،)]+",
};
Expand Down
6 changes: 6 additions & 0 deletions src/llama.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5443,6 +5443,12 @@ static void llm_load_vocab(
} else if (
tokenizer_pre == "codeshell") {
vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_CODESHELL;
} else if (
tokenizer_pre == "bloom") {
vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_BLOOM;
} else if (
tokenizer_pre == "gpt3-finnish") {
vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_GPT3_FINNISH;
} else {
throw std::runtime_error(format("unknown pre-tokenizer type: '%s'", tokenizer_pre.c_str()));
}
Expand Down
Loading