Skip to content

Local UI to run and train LLMs and diffusion models. Supports GGUF, MLX, Qwen3.8, DeepSeek-V4.1, MiniMax-H3, Gemma 4, FLUX and more.

5.2 KB
Raw Blame
"""Works on the inner tokenizer of a multimodal processor, since MLX processors lack padding_side."""

import ast
import os

CHAT_TEMPLATES_PATH = os.path.join(
    os.path.dirname(os.path.dirname(os.path.abspath(__file__))),
    "unsloth",
    "chat_templates.py",
)

_BODY = next(
    node.body
    for node in ast.parse(open(CHAT_TEMPLATES_PATH, encoding = "utf-8").read()).body
    if isinstance(node, ast.FunctionDef) and node.name == "get_chat_template"
)


def _only(found, what):
    assert len(found) == 1, f"could not find {what} in get_chat_template"
    return found[0]


def _assign(name):
    """The one top-level `<name> = ...` in get_chat_template."""
    return _only(
        [
            node
            for node in _BODY
            if isinstance(node, ast.Assign)
            and any(getattr(target, "id", None) == name for target in node.targets)
        ],
        f"the {name} binding",
    )


def _unwrap_branch():
    """The top-level `if ...` whose body claims the processor."""
    return _only(
        [
            node
            for node in _BODY
            if isinstance(node, ast.If)
            and any(
                isinstance(stmt, ast.Assign)
                and any(getattr(target, "id", None) == "_processor" for target in stmt.targets)
                for stmt in node.body
            )
        ],
        "the processor unwrap",
    )


def _reattach_branch():
    """The top-level `if _processor is not None:` that restores it before return."""
    return _only(
        [
            node
            for node in _BODY
            if isinstance(node, ast.If) and "_processor" in ast.dump(node.test)
        ],
        "the processor re-attach",
    )


def _run(statements, **namespace):
    namespace.setdefault("ProcessorMixin", _ProcessorMixin)
    exec(
        compile(ast.Module(body = statements, type_ignores = []), CHAT_TEMPLATES_PATH, "exec"),
        namespace,
    )
    return namespace


class _ProcessorMixin:
    pass


class _FakeTokenizer:
    """A tokenizer as this function sees one: special tokens, no `.tokenizer`."""

    def __init__(
        self,
        eos_token = "<eos>",
        eos_token_id = 1,
    ):
        self.padding_side = "right"
        self.bos_token, self.bos_token_id = "<bos>", 0
        self.eos_token, self.eos_token_id = eos_token, eos_token_id
        self.pad_token, self.pad_token_id = "<pad>", 2


class _FakeProcessor(_ProcessorMixin):
    """A processor as this function sees one: a `.tokenizer`, plus the loader's mirror."""

    def __init__(self, tokenizer):
        self.tokenizer = tokenizer
        for token in ("bos_token", "eos_token", "pad_token"):
            setattr(self, token, getattr(tokenizer, token))
            setattr(self, token + "_id", getattr(tokenizer, token + "_id"))


class _FakeTokenizerBackend(_FakeTokenizer):
    """A tokenizer backend may expose `.tokenizer` without being a processor."""

    def __init__(self):
        super().__init__()
        self.tokenizer = object()


def _unwrap(tokenizer):
    return _run(
        [_assign("_processor"), _unwrap_branch(), _assign("old_tokenizer")],
        tokenizer = tokenizer,
    )


def _reattach(processor, tokenizer):
    return _run(
        [_reattach_branch()],
        _processor = processor,
        tokenizer = tokenizer,
        chat_template = "TEMPLATE",
        ollama_modelfile = "MODELFILE",
        ollama_system_message = None,
    )


def test_processor_is_unwrapped_to_its_inner_tokenizer():
    tokenizer = _FakeTokenizer()
    namespace = _unwrap(_FakeProcessor(tokenizer))
    assert namespace["tokenizer"] is tokenizer
    assert namespace["tokenizer"].padding_side == "right"


def test_old_tokenizer_is_the_inner_tokenizer_not_the_processor():
    # pad_token is None on the MLX processor, so binding old_tokenizer before the unwrap blanks tokens.
    tokenizer = _FakeTokenizer()
    assert _unwrap(_FakeProcessor(tokenizer))["old_tokenizer"] is tokenizer


def test_a_plain_tokenizer_is_left_alone():
    tokenizer = _FakeTokenizer()
    namespace = _unwrap(tokenizer)
    assert namespace["_processor"] is None
    assert namespace["tokenizer"] is tokenizer
    assert namespace["old_tokenizer"] is tokenizer


def test_a_tokenizer_backend_with_a_tokenizer_attribute_is_left_alone():
    tokenizer = _FakeTokenizerBackend()
    namespace = _unwrap(tokenizer)
    assert namespace["_processor"] is None
    assert namespace["tokenizer"] is tokenizer
    assert namespace["old_tokenizer"] is tokenizer


def test_processor_is_returned_carrying_the_new_tokenizer_and_template():
    processor = _FakeProcessor(_FakeTokenizer())
    rebuilt = _FakeTokenizer()  # the vocab surgery replaces the tokenizer object
    namespace = _reattach(processor, rebuilt)
    assert namespace["tokenizer"] is processor
    assert processor.tokenizer is rebuilt
    assert processor.chat_template == "TEMPLATE"


def test_a_remapped_eos_reaches_the_processors_mirrored_copy():
    # chatml/gemma_chatml remap eos; the collators and saves read the processor's copy, so it must follow.
    processor = _FakeProcessor(_FakeTokenizer(eos_token = "<eos>", eos_token_id = 1))
    rebuilt = _FakeTokenizer(eos_token = "<|im_end|>", eos_token_id = 107)
    _reattach(processor, rebuilt)
    assert processor.eos_token == "<|im_end|>"
    assert processor.eos_token_id == 107