Dataset¶

multimolecule.data.Dataset ¶

Bases: Dataset

The base class for all datasets.

Dataset is a subclass of datasets.Dataset that provides additional functionality for handling structured data. It has three main features:

column identification: identify the special columns (sequence and structure columns) in the dataset.
tokenization: tokenize the sequence columns in the dataset using a pretrained tokenizer.
task inference: infer the task type and level of each label column in the dataset.

Attributes:

Name	Type	Description
`task`		A [`Task`][] object that describes the task type and level of the label column in the dataset.
`tokenizer`	`PreTrainedTokenizerBase`	The pretrained tokenizer to use for tokenization.
`truncation`	`bool`	Whether to truncate sequences that exceed the maximum length of the tokenizer.
`max_seq_length`	`int`	The maximum length of the input sequences.
`data_cols`	`List`	The names of all columns in the dataset.
`feature_cols`	`List`	The names of the feature columns in the dataset.
`label_col`	`str`	The names of the label column in the dataset.
`sequence_col`	`str`	The names of the sequence column in the dataset.
`column_names_map`	`Mapping[str, str] \| None`	A mapping of column names to new column names.
`preprocess`	`bool`	Whether to preprocess the dataset.

Parameters:

Name	Type	Description	Default
`data` ¶	`Table \| DataFrame \| dict \| list \| str`	The dataset. This can be a path to a file, a tag on the Hugging Face Hub, a pyarrow.Table, a dict, a list, or a pandas.DataFrame.	required
`split` ¶	`NamedSplit \| None`	The split of the dataset.	`None`
`tokenizer` ¶	`PreTrainedTokenizerBase \| None`	A pretrained tokenizer to use for tokenization. Either `tokenizer` or `pretrained` must be specified.	`None`
`pretrained` ¶	`str \| None`	The name of a pretrained tokenizer to use for tokenization. Either `tokenizer` or `pretrained` must be specified.	`None`
`feature_cols` ¶	`List \| None`	The names of the feature columns in the dataset. Will be inferred automatically if not specified.	`None`
`label_col` ¶	`str \| None`	The names of the label column in the dataset. Will be inferred automatically if not specified.	`None`
`id_cols` ¶	`List \| None`	The names of the ID columns in the dataset. Will be inferred automatically if not specified.	`None`
`preprocess` ¶	`bool \| None`	Whether to preprocess the dataset. Preprocessing involves pre-tokenizing the sequences using the tokenizer. Defaults to `True`.	`None`
`auto_rename_sequence_col` ¶	`bool \| None`	Whether to automatically rename sequence columns to standard name. Only works when there is exactly one sequence column You can control the naming through `multimolecule.defaults.SEQUENCE_COL_NAME`. For more refined control, use `column_names_map`.	`None`
`auto_rename_label_col` ¶	`bool \| None`	Whether to automatically rename label column to standard name. Only works when there is exactly one label column. You can control the naming through `multimolecule.defaults.LABEL_COL_NAME`. For more refined control, use `column_names_map`.	`None`
`column_names_map` ¶	`Mapping[str, str] \| None`	A mapping of column names to new column names. This is useful for renaming columns to inputs that are expected by a model. Defaults to `None`.	`None`
`truncation` ¶	`bool \| None`	Whether to truncate sequences that exceed the maximum length of the tokenizer. Defaults to `False`.	`None`
`max_seq_length` ¶	`int \| None`	The maximum length of the input sequences. Defaults to the `model_max_length` of the tokenizer.	`None`
`task` ¶	`Task \| None`	A [`Task`][] object that describes the task type and level of the label column in the dataset. Will be inferred automatically if not specified.	`None`
`discrete_map` ¶	`Mapping[str, int] \| None`	A mapping of column names to discrete mappings. This is useful for mapping the raw value to nominal value in classification task. Will be inferred automatically if not specified.	`None`
`nan_process` ¶	`str`	How to handle NaN and inf values in the dataset. Can be “ignore”, “error”, “drop”, or “fill”. Defaults to “ignore”.	`'ignore'`
`fill_value` ¶	`str \| int \| float`	The value to fill NaN and inf values with. Defaults to 0.	`0`
`info` ¶	`DatasetInfo \| None`	The dataset info.	`None`
`indices_table` ¶		The indices table.	required
`fingerprint` ¶	`str \| None`	The fingerprint of the dataset.	`None`

Source code in multimolecule/data/dataset.py

Python
@DatasetRegistry.register("auto", default=True)
class Dataset(datasets.Dataset):
    r"""
    The base class for all datasets.

    Dataset is a subclass of [`datasets.Dataset`][] that provides additional functionality for handling structured data.
    It has three main features:

    - column identification: identify the special columns (sequence and structure columns) in the dataset.
    - tokenization: tokenize the sequence columns in the dataset using a pretrained tokenizer.
    - task inference: infer the task type and level of each label column in the dataset.

    Attributes:
        task: A [`Task`][] object that describes the task type and level of the label column in the dataset.
        tokenizer: The pretrained tokenizer to use for tokenization.
        truncation: Whether to truncate sequences that exceed the maximum length of the tokenizer.
        max_seq_length: The maximum length of the input sequences.
        data_cols: The names of all columns in the dataset.
        feature_cols: The names of the feature columns in the dataset.
        label_col: The names of the label column in the dataset.
        sequence_col: The names of the sequence column in the dataset.
        column_names_map: A mapping of column names to new column names.
        preprocess: Whether to preprocess the dataset.

    Args:
        data: The dataset. This can be a path to a file, a tag on the Hugging Face Hub, a pyarrow.Table,
            a [dict][], a [list][], or a [pandas.DataFrame][].
        split: The split of the dataset.
        tokenizer: A pretrained tokenizer to use for tokenization.
            Either `tokenizer` or `pretrained` must be specified.
        pretrained: The name of a pretrained tokenizer to use for tokenization.
            Either `tokenizer` or `pretrained` must be specified.
        feature_cols: The names of the feature columns in the dataset.
            Will be inferred automatically if not specified.
        label_col: The names of the label column in the dataset.
            Will be inferred automatically if not specified.
        id_cols: The names of the ID columns in the dataset.
            Will be inferred automatically if not specified.
        preprocess: Whether to preprocess the dataset.
            Preprocessing involves pre-tokenizing the sequences using the tokenizer.
            Defaults to `True`.
        auto_rename_sequence_col: Whether to automatically rename sequence columns to standard name.
            Only works when there is exactly one sequence column
            You can control the naming through `multimolecule.defaults.SEQUENCE_COL_NAME`.
            For more refined control, use `column_names_map`.
        auto_rename_label_col: Whether to automatically rename label column to standard name.
            Only works when there is exactly one label column.
            You can control the naming through `multimolecule.defaults.LABEL_COL_NAME`.
            For more refined control, use `column_names_map`.
        column_names_map: A mapping of column names to new column names.
            This is useful for renaming columns to inputs that are expected by a model.
            Defaults to `None`.
        truncation: Whether to truncate sequences that exceed the maximum length of the tokenizer.
            Defaults to `False`.
        max_seq_length: The maximum length of the input sequences.
            Defaults to the `model_max_length` of the tokenizer.
        task: A [`Task`][] object that describes the task type and level of the label column in the dataset.
            Will be inferred automatically if not specified.
        discrete_map: A mapping of column names to discrete mappings.
            This is useful for mapping the raw value to nominal value in classification task.
            Will be inferred automatically if not specified.
        nan_process: How to handle NaN and inf values in the dataset.
            Can be "ignore", "error", "drop", or "fill". Defaults to "ignore".
        fill_value: The value to fill NaN and inf values with.
            Defaults to 0.
        info: The dataset info.
        indices_table: The indices table.
        fingerprint: The fingerprint of the dataset.
    """

    _id_cols: List[str]
    _feature_cols: List[str]
    _label_col: str

    _sequence_col: str
    _sequence_type: str
    _secondary_structure_cols: List[str]

    _task: Task
    _discrete_map: Mapping[str, Any]

    tokenizer: PreTrainedTokenizerBase
    truncation: bool = False
    max_seq_length: int
    seq_length_offset: int = 0

    preprocess: bool = True
    auto_rename_sequence_col: bool = True
    auto_rename_label_col: bool = True
    column_names_map: Mapping[str, str] | None = None
    ignored_cols: List[str] = []

    def __init__(
        self,
        data: Table | DataFrame | dict | list | str,
        split: datasets.NamedSplit | None = None,
        feature_cols: List | None = None,
        label_col: str | None = None,
        id_cols: List | None = None,
        sequence_col: str | None = None,
        ignored_cols: List | None = None,
        auto_rename_sequence_col: bool | None = None,
        auto_rename_label_col: bool | None = None,
        column_names_map: Mapping[str, str] | None = None,
        task: Task | None = None,
        discrete_map: Mapping[str, int] | None = None,
        sequence_type: str | None = None,
        tokenizer: PreTrainedTokenizerBase | None = None,
        pretrained: str | None = None,
        truncation: bool | None = None,
        max_seq_length: int | None = None,
        nan_process: str = "ignore",
        fill_value: str | int | float = 0,
        info: datasets.DatasetInfo | None = None,
        indices: Sequence | Table | None = None,
        fingerprint: str | None = None,
        preprocess: bool | None = None,
        train: bool | None = None,
    ):
        arrow_table = self.build_table(
            data, split, feature_cols, label_col, nan_process=nan_process, fill_value=fill_value
        )
        if indices is not None and not isinstance(indices, (Table, pa.Table)):
            indices = pa.Table.from_arrays([pa.array(indices, type=pa.uint32())], names=["index"])
        super().__init__(
            arrow_table=arrow_table, split=split, info=info, indices_table=indices, fingerprint=fingerprint
        )
        self.post_init(
            feature_cols=feature_cols,
            label_col=label_col,
            id_cols=id_cols,
            sequence_col=sequence_col,
            ignored_cols=ignored_cols,
            task=task,
            discrete_map=discrete_map,
            auto_rename_sequence_col=auto_rename_sequence_col,
            auto_rename_label_col=auto_rename_label_col,
            column_names_map=column_names_map,
            sequence_type=sequence_type,
            tokenizer=tokenizer,
            pretrained=pretrained,
            truncation=truncation,
            max_seq_length=max_seq_length,
            preprocess=preprocess,
            train=train,
        )

    def build_table(
        self,
        data: Table | DataFrame | dict | str,
        split: datasets.NamedSplit,
        feature_cols: List | None = None,
        label_col: str | None = None,
        nan_process: str | None = "ignore",
        fill_value: str | int | float = 0,
    ) -> datasets.table.Table:
        if isinstance(data, str):
            try:
                data = datasets.load_dataset(data, split=split).data
            except (FileNotFoundError, ValueError):
                data = dl.load_pandas(data)
        if isinstance(data, dict):
            data = pa.Table.from_pydict(data)
        elif isinstance(data, list):
            data = pa.Table.from_pylist(data)
        elif isinstance(data, DataFrame):
            data = data.loc[:, ~data.columns.str.contains("^Unnamed")]
            # If there are None values in the dataset, we replace them with nan and process them later at once.
            data = data.fillna(float("nan"))
            pd_version = parse_version(pd.__version__)
            if pd_version >= parse_version("2.1.0"):
                data = data.map(lambda x: [float("nan") if i is None else i for i in x] if isinstance(x, list) else x)
            else:
                data = data.applymap(
                    lambda x: [float("nan") if i is None else i for i in x] if isinstance(x, list) else x
                )
            data = pa.Table.from_pandas(data, preserve_index=False)
        if feature_cols is not None and label_col is not None:
            data = data.select(feature_cols + [label_col])
        data = self.process_nan(data, nan_process=nan_process, fill_value=fill_value)
        return data

    def post_init(
        self,
        feature_cols: List | None = None,
        label_col: str | None = None,
        id_cols: List | None = None,
        sequence_col: str | None = None,
        ignored_cols: List | None = None,
        task: Task | None = None,
        discrete_map: Mapping[str, int] | None = None,
        auto_rename_sequence_col: bool | None = None,
        auto_rename_label_col: bool | None = None,
        column_names_map: Mapping[str, str] | None = None,
        sequence_type: str | None = None,
        tokenizer: PreTrainedTokenizerBase | None = None,
        pretrained: str | None = None,
        max_seq_length: int | None = None,
        truncation: bool | None = None,
        preprocess: bool | None = None,
        train: bool | None = None,
    ) -> None:
        r"""
        Perform pre-processing steps after initialization.

        It first identifies the special columns (sequence and structure columns) in the dataset.
        Then it sets the feature and label columns based on the input arguments.
        If `auto_rename_sequence_col` is `True`, it will automatically rename the sequence column.
        If `auto_rename_label_col` is `True`, it will automatically rename the label column.
        Finally, it sets the [`transform`][datasets.Dataset.set_transform] function based on the `preprocess` flag.
        """

        # Process columns
        self.identify_special_cols(
            feature_cols=feature_cols,
            label_col=label_col,
            id_cols=id_cols,
            sequence_col=sequence_col,
            sequence_type=sequence_type,
        )
        self.ignored_cols = ignored_cols or self.id_cols
        if auto_rename_sequence_col is not None:
            self.auto_rename_sequence_col = auto_rename_sequence_col
        if auto_rename_label_col is not None:
            self.auto_rename_label_col = auto_rename_label_col
        if column_names_map is None:
            column_names_map = {}
        if self.auto_rename_sequence_col:
            column_names_map[self.sequence_col] = defaults.SEQUENCE_COL_NAME  # type: ignore[index]
        if self.auto_rename_label_col:
            column_names_map[self.label_col] = defaults.LABEL_COL_NAME  # type: ignore[index]
        self.column_names_map = column_names_map
        if self.column_names_map:
            self.rename_columns(self.column_names_map)

        # Initialize tokenizer
        if tokenizer is None:
            if not pretrained:
                if not self.sequence_type:
                    raise ValueError("Either pretrained or sequence_type must be specified.")
                pretrained = "multimolecule/" + self.sequence_type.lower()
            tokenizer = AutoTokenizer.from_pretrained(pretrained)
        if max_seq_length is None:
            max_seq_length = tokenizer.model_max_length
        else:
            tokenizer.model_max_length = max_seq_length
        self.tokenizer = tokenizer
        self.max_seq_length = max_seq_length
        if truncation is not None:
            self.truncation = truncation
        if self.tokenizer.cls_token is not None:
            self.seq_length_offset += 1
        if self.tokenizer.sep_token is not None and self.tokenizer.sep_token != self.tokenizer.eos_token:
            self.seq_length_offset += 1
        if self.tokenizer.eos_token is not None:
            self.seq_length_offset += 1

        # Infer Tasks & Discrete Map
        if task is not None:
            self.task = task
        else:
            self.task = self.infer_task()
        if discrete_map is not None:
            self._discrete_map = discrete_map
        self.train = train if train is not None else self.split.lower() in defaults.TRAIN_SPLITS

        # Preprocess
        if preprocess is not None:
            self.preprocess = preprocess
        if self.preprocess:
            self.update(self.map(self.tokenization))
            if self.secondary_structure_cols:
                self.update(self.map(self.convert_secondary_structure))
            if self.discrete_map:
                self.update(self.map(self.map_discrete))
            if self.truncation and 0 < self.max_seq_length < 2**32:
                max_seq_length = self.max_seq_length - self.seq_length_offset
                self.update(self.map(self.truncate, fn_kwargs={"max_seq_length": max_seq_length}))
        self.set_transform(self.transform)

    def transform(self, batch: Mapping) -> Mapping:
        r"""
        Default [`transform`][datasets.Dataset.set_transform].

        See Also:
            [`collate`][multimolecule.Dataset.collate]
        """
        return {k: self.collate(k, v) for k, v in batch.items() if k not in self.ignored_cols}

    def collate(self, col: str, data: Any) -> Tensor | NestedTensor | None:
        r"""
        Collate the data for a column.

        If the column is a sequence column, it will tokenize the data if `tokenize` is `True`.
        Otherwise, it will return a tensor or nested tensor.
        """
        if col == self.sequence_col:
            if isinstance(data[0], str):
                data = self.tokenize(data)
            return NestedTensor(data)
        if not self.preprocess:
            if col in self.discrete_map:
                data = map_value(data, self.discrete_map[col])
            if col == self.label_col:
                data = truncate_value(data, self.max_seq_length - self.seq_length_offset, self.task.level)
        if col == self.label_col:
            ignore_value = float("nan") if self.task.type == TaskType.Regression else -100
            if isinstance(data[0], list):
                data = [[i if i is not None else ignore_value for i in d] for d in data]
            else:
                data = [i if i is not None else ignore_value for i in data]
        if isinstance(data[0], str):
            return data
        try:
            return torch.tensor(data)
        except ValueError:
            return NestedTensor(data)

    def infer_task(self, sequence_col: str | None = None) -> Task:
        if sequence_col is None:
            sequence_col = self.sequence_col
        if self.label_col in self.secondary_structure_cols:
            task = Task(TaskType.Binary, level=TaskLevel.Contact, num_labels=1)
            warn(
                f"Secondary Structure task is assumed to be {task}. "
                "Please explicitly specify the task if this is not the case."
            )
            return task
        sequence = self._data.column(sequence_col)
        column = self._data.column(self.label_col)
        return infer_task(
            sequence,
            column,
            truncation=self.truncation,
            max_seq_length=self.max_seq_length,
            seq_length_offset=self.seq_length_offset,
        )

    def infer_discrete_map(self, discrete_map: Mapping | None = None):
        self._discrete_map = discrete_map or NestedDict()
        ignored_cols = set(self.discrete_map.keys()) | set(self.sequence_col) | set(self.secondary_structure_cols)
        data_cols = [i for i in self.data_cols if i not in ignored_cols]
        for col in data_cols:
            discrete_map = infer_discrete_map(self._data.column(col))
            if discrete_map:
                self._discrete_map[col] = discrete_map  # type: ignore[index]
        return self._discrete_map

    def __getitems__(self, keys: int | slice | Iterable[int]) -> Any:
        return self.__getitem__(keys)

    def identify_special_cols(
        self,
        feature_cols: List | None = None,
        label_col: str | None = None,
        id_cols: List | None = None,
        sequence_col: str | None = None,
        sequence_type: str | None = None,
    ) -> Sequence:
        all_cols = self.data.column_names
        self._id_cols = id_cols or [i for i in all_cols if i.lower() in defaults.ID_COL_NAMES]
        self._sequence_col = sequence_col  # type: ignore[assignment]
        self._sequence_type = sequence_type  # type: ignore[assignment]

        string_cols: list[str] = [k for k, v in self.features.items() if k not in self.id_cols and v.dtype == "string"]
        unique_chars = {
            k: {ch for s in flatten_column(self._data.column(k))[0] for ch in s.as_py()} for k in string_cols
        }
        unique_chars_upper = {k: {ch.upper() for ch in v} for k, v in unique_chars.items()}

        if self._sequence_col is None:
            break_flag = False
            for col, chars in unique_chars_upper.items():
                for alphabet_type, alphabet in alphabets.items():
                    complete, minimal = alphabet["complete"], alphabet["minimal"]
                    if chars.issubset(complete) and chars.issuperset(minimal):
                        self._sequence_col = col
                        if self._sequence_type is None and alphabet_type != "na":
                            self._sequence_type = alphabet_type
                        break_flag = True
                        break
                if break_flag:
                    break
            else:
                raise ValueError("No sequence column found in the dataset.")

        self._secondary_structure_cols = [
            k for k, v in unique_chars.items() if v.issubset(DB_COMPLETE_ALPHABET) and v.issuperset(DB_MINIMAL_ALPHABET)
        ]

        data_cols = [i for i in all_cols if i not in self.id_cols]
        if label_col is None:
            if feature_cols is None:
                feature_cols = [i for i in data_cols if i in defaults.SEQUENCE_COL_NAMES]
            label_col = [i for i in data_cols if i not in feature_cols][0]
        self._label_col = label_col
        if feature_cols is None:
            feature_cols = [i for i in data_cols if i not in self.label_col]
        self._feature_cols = feature_cols
        missing_feature_cols = set(self.feature_cols).difference(data_cols)
        if missing_feature_cols:
            raise ValueError(f"{missing_feature_cols} are specified in feature_cols, but not found in dataset.")
        missing_label_col = label_col not in data_cols
        if missing_label_col:
            raise ValueError(f"{label_col} is specified as label_col, but not found in dataset.")
        return string_cols

    def tokenize(self, string: str) -> Tensor:
        return self.tokenizer(string, return_attention_mask=False, truncation=self.truncation)["input_ids"]

    def tokenization(self, data: Mapping[str, str]) -> Mapping[str, Tensor]:
        return {self.sequence_col: self.tokenize(data[self.sequence_col])}

    def convert_secondary_structure(self, data: Mapping) -> Mapping:
        return {col: dot_bracket_to_contact_map(data[col]) for col in self.secondary_structure_cols}

    def map_discrete(self, data: Mapping) -> Mapping:
        return {name: map_value(data[name], mapping) for name, mapping in self.discrete_map.items()}

    def truncate(self, data: Mapping, max_seq_length: int) -> Mapping:
        return {self.label_col: truncate_value(data[self.label_col], max_seq_length, self.task.level)}

    def update(self, dataset: datasets.Dataset):
        r"""
        Perform an in-place update of the dataset.

        This method is used to update the dataset after changes have been made to the underlying data.
        It updates the format columns, data, info, and fingerprint of the dataset.
        """
        # pylint: disable=W0212
        # Why datasets won't support in-place changes?
        # It's just impossible to extend.
        self._format_columns = dataset._format_columns
        self._data = dataset._data
        self._info = dataset._info
        self._fingerprint = dataset._fingerprint

    def rename_columns(self, column_mapping: Mapping[str, str], new_fingerprint: str | None = None) -> datasets.Dataset:
        self.update(super().rename_columns(column_mapping, new_fingerprint=new_fingerprint))
        self._id_cols = [column_mapping.get(i, i) for i in self.id_cols]
        self._feature_cols = [column_mapping.get(i, i) for i in self.feature_cols]
        if self._label_col in column_mapping:
            self._label_col = column_mapping[self.label_col]
        if self._sequence_col in column_mapping:
            self._sequence_col = column_mapping[self.sequence_col]
        self._secondary_structure_cols = [column_mapping.get(i, i) for i in self.secondary_structure_cols]
        return self

    def rename_column(
        self, original_column_name: str, new_column_name: str, new_fingerprint: str | None = None
    ) -> datasets.Dataset:
        self.update(super().rename_column(original_column_name, new_column_name, new_fingerprint))
        self._id_cols = [new_column_name if i == original_column_name else i for i in self.id_cols]
        self._feature_cols = [new_column_name if i == original_column_name else i for i in self.feature_cols]
        if self._label_col == original_column_name:
            self._label_col = new_column_name
        if self._sequence_col == original_column_name:
            self._sequence_col = new_column_name
        self._secondary_structure_cols = [
            new_column_name if i == original_column_name else i for i in self.secondary_structure_cols
        ]
        return self

    def process_nan(self, data: Table, nan_process: str | None, fill_value: str | int | float = 0) -> Table:
        if nan_process == "ignore":
            return data
        data = data.to_pandas()
        data = data.replace([float("inf"), -float("inf")], float("nan"))
        if data.isnull().values.any():
            if nan_process is None or nan_process == "error":
                raise ValueError("NaN / inf values have been found in the dataset.")
            warn(
                "NaN / inf values have been found in the dataset.\n"
                "While we can handle them, the data type of the corresponding column may be set to float, "
                "which can and very likely will disrupt the auto task recognition.\n"
                "It is recommended to address these values before loading the dataset."
            )
            if nan_process == "drop":
                data = data.dropna()
            elif nan_process == "fill":
                data = data.fillna(fill_value)
            else:
                raise ValueError(f"Invalid nan_process: {nan_process}")
        return pa.Table.from_pandas(data, preserve_index=False)

    @property
    def id_cols(self) -> List:
        return self._id_cols

    @property
    def data_cols(self) -> List:
        return self.feature_cols + [self.label_col]

    @property
    def feature_cols(self) -> List:
        return self._feature_cols

    @property
    def label_col(self) -> str:
        return self._label_col

    @property
    def sequence_col(self) -> str:
        return self._sequence_col

    @property
    def secondary_structure_cols(self) -> List:
        return self._secondary_structure_cols

    @property
    def sequence_type(self) -> str:
        return self._sequence_type

    @property
    def discrete_map(self) -> Mapping:
        if not hasattr(self, "_discrete_map"):
            return self.infer_discrete_map()
        return self._discrete_map

post_init ¶

Python

post_init(feature_cols: List | None = None, label_col: str | None = None, id_cols: List | None = None, sequence_col: str | None = None, ignored_cols: List | None = None, task: Task | None = None, discrete_map: Mapping[str, int] | None = None, auto_rename_sequence_col: bool | None = None, auto_rename_label_col: bool | None = None, column_names_map: Mapping[str, str] | None = None, sequence_type: str | None = None, tokenizer: PreTrainedTokenizerBase | None = None, pretrained: str | None = None, max_seq_length: int | None = None, truncation: bool | None = None, preprocess: bool | None = None, train: bool | None = None) -> None

Perform pre-processing steps after initialization.

It first identifies the special columns (sequence and structure columns) in the dataset. Then it sets the feature and label columns based on the input arguments. If auto_rename_sequence_col is True, it will automatically rename the sequence column. If auto_rename_label_col is True, it will automatically rename the label column. Finally, it sets the transform function based on the preprocess flag.

Source code in multimolecule/data/dataset.py

Python
def post_init(
    self,
    feature_cols: List | None = None,
    label_col: str | None = None,
    id_cols: List | None = None,
    sequence_col: str | None = None,
    ignored_cols: List | None = None,
    task: Task | None = None,
    discrete_map: Mapping[str, int] | None = None,
    auto_rename_sequence_col: bool | None = None,
    auto_rename_label_col: bool | None = None,
    column_names_map: Mapping[str, str] | None = None,
    sequence_type: str | None = None,
    tokenizer: PreTrainedTokenizerBase | None = None,
    pretrained: str | None = None,
    max_seq_length: int | None = None,
    truncation: bool | None = None,
    preprocess: bool | None = None,
    train: bool | None = None,
) -> None:
    r"""
    Perform pre-processing steps after initialization.

    It first identifies the special columns (sequence and structure columns) in the dataset.
    Then it sets the feature and label columns based on the input arguments.
    If `auto_rename_sequence_col` is `True`, it will automatically rename the sequence column.
    If `auto_rename_label_col` is `True`, it will automatically rename the label column.
    Finally, it sets the [`transform`][datasets.Dataset.set_transform] function based on the `preprocess` flag.
    """

    # Process columns
    self.identify_special_cols(
        feature_cols=feature_cols,
        label_col=label_col,
        id_cols=id_cols,
        sequence_col=sequence_col,
        sequence_type=sequence_type,
    )
    self.ignored_cols = ignored_cols or self.id_cols
    if auto_rename_sequence_col is not None:
        self.auto_rename_sequence_col = auto_rename_sequence_col
    if auto_rename_label_col is not None:
        self.auto_rename_label_col = auto_rename_label_col
    if column_names_map is None:
        column_names_map = {}
    if self.auto_rename_sequence_col:
        column_names_map[self.sequence_col] = defaults.SEQUENCE_COL_NAME  # type: ignore[index]
    if self.auto_rename_label_col:
        column_names_map[self.label_col] = defaults.LABEL_COL_NAME  # type: ignore[index]
    self.column_names_map = column_names_map
    if self.column_names_map:
        self.rename_columns(self.column_names_map)

    # Initialize tokenizer
    if tokenizer is None:
        if not pretrained:
            if not self.sequence_type:
                raise ValueError("Either pretrained or sequence_type must be specified.")
            pretrained = "multimolecule/" + self.sequence_type.lower()
        tokenizer = AutoTokenizer.from_pretrained(pretrained)
    if max_seq_length is None:
        max_seq_length = tokenizer.model_max_length
    else:
        tokenizer.model_max_length = max_seq_length
    self.tokenizer = tokenizer
    self.max_seq_length = max_seq_length
    if truncation is not None:
        self.truncation = truncation
    if self.tokenizer.cls_token is not None:
        self.seq_length_offset += 1
    if self.tokenizer.sep_token is not None and self.tokenizer.sep_token != self.tokenizer.eos_token:
        self.seq_length_offset += 1
    if self.tokenizer.eos_token is not None:
        self.seq_length_offset += 1

    # Infer Tasks & Discrete Map
    if task is not None:
        self.task = task
    else:
        self.task = self.infer_task()
    if discrete_map is not None:
        self._discrete_map = discrete_map
    self.train = train if train is not None else self.split.lower() in defaults.TRAIN_SPLITS

    # Preprocess
    if preprocess is not None:
        self.preprocess = preprocess
    if self.preprocess:
        self.update(self.map(self.tokenization))
        if self.secondary_structure_cols:
            self.update(self.map(self.convert_secondary_structure))
        if self.discrete_map:
            self.update(self.map(self.map_discrete))
        if self.truncation and 0 < self.max_seq_length < 2**32:
            max_seq_length = self.max_seq_length - self.seq_length_offset
            self.update(self.map(self.truncate, fn_kwargs={"max_seq_length": max_seq_length}))
    self.set_transform(self.transform)

transform ¶

Python

transform(batch: Mapping) -> Mapping

Default transform.

collate ¶

Python

collate(col: str, data: Any) -> Tensor | NestedTensor | None

Collate the data for a column.

If the column is a sequence column, it will tokenize the data if tokenize is True. Otherwise, it will return a tensor or nested tensor.

Source code in multimolecule/data/dataset.py

Python
def collate(self, col: str, data: Any) -> Tensor | NestedTensor | None:
    r"""
    Collate the data for a column.

    If the column is a sequence column, it will tokenize the data if `tokenize` is `True`.
    Otherwise, it will return a tensor or nested tensor.
    """
    if col == self.sequence_col:
        if isinstance(data[0], str):
            data = self.tokenize(data)
        return NestedTensor(data)
    if not self.preprocess:
        if col in self.discrete_map:
            data = map_value(data, self.discrete_map[col])
        if col == self.label_col:
            data = truncate_value(data, self.max_seq_length - self.seq_length_offset, self.task.level)
    if col == self.label_col:
        ignore_value = float("nan") if self.task.type == TaskType.Regression else -100
        if isinstance(data[0], list):
            data = [[i if i is not None else ignore_value for i in d] for d in data]
        else:
            data = [i if i is not None else ignore_value for i in data]
    if isinstance(data[0], str):
        return data
    try:
        return torch.tensor(data)
    except ValueError:
        return NestedTensor(data)

update ¶

Python

update(dataset: Dataset)

Perform an in-place update of the dataset.

This method is used to update the dataset after changes have been made to the underlying data. It updates the format columns, data, info, and fingerprint of the dataset.

Source code in multimolecule/data/dataset.py

Python
def update(self, dataset: datasets.Dataset):
    r"""
    Perform an in-place update of the dataset.

    This method is used to update the dataset after changes have been made to the underlying data.
    It updates the format columns, data, info, and fingerprint of the dataset.
    """
    # pylint: disable=W0212
    # Why datasets won't support in-place changes?
    # It's just impossible to extend.
    self._format_columns = dataset._format_columns
    self._data = dataset._data
    self._info = dataset._info
    self._fingerprint = dataset._fingerprint

Dataset¶

multimolecule.data.Dataset ¶

`data` ¶

`split` ¶

`tokenizer` ¶

`pretrained` ¶

`feature_cols` ¶

`label_col` ¶

`id_cols` ¶

`preprocess` ¶

`auto_rename_sequence_col` ¶

`auto_rename_label_col` ¶

`column_names_map` ¶

`truncation` ¶

`max_seq_length` ¶

`task` ¶

`discrete_map` ¶

`nan_process` ¶

`fill_value` ¶

`info` ¶

`indices_table` ¶

`fingerprint` ¶

post_init ¶

transform ¶

collate ¶

update ¶

Dataset¶

multimolecule.data.Dataset ¶

data ¶

split ¶

tokenizer ¶

pretrained ¶

feature_cols ¶

label_col ¶

id_cols ¶

preprocess ¶

auto_rename_sequence_col ¶

auto_rename_label_col ¶

column_names_map ¶

truncation ¶

max_seq_length ¶

task ¶

discrete_map ¶

nan_process ¶

fill_value ¶

info ¶

indices_table ¶

fingerprint ¶

post_init ¶

transform ¶

collate ¶

update ¶

`data` ¶

`split` ¶

`tokenizer` ¶

`pretrained` ¶

`feature_cols` ¶

`label_col` ¶

`id_cols` ¶

`preprocess` ¶

`auto_rename_sequence_col` ¶

`auto_rename_label_col` ¶

`column_names_map` ¶

`truncation` ¶

`max_seq_length` ¶

`task` ¶

`discrete_map` ¶

`nan_process` ¶

`fill_value` ¶

`info` ¶

`indices_table` ¶

`fingerprint` ¶