data_juicer.ops

data_juicer.ops.load_ops(process_list, op_env_manager=None)

Load op list according to the process list from config file.

参数:

• process_list -- A process list. Each item is an op name and its arguments.

• op_env_manager -- The OPEnvManager to try to merge environment specs of different OPs that have common dependencies. Only available when min_common_dep_num_to_combine >= 0.

返回:

The op instance list.

class data_juicer.ops.Filter(*args, **kwargs)

基类：OP

__init__(*args, **kwargs)

Base class that removes specific info.

参数:

• text_key -- the key name of field that stores sample texts to be processed

• image_key -- the key name of field that stores sample image list to be processed

• audio_key -- the key name of field that stores sample audio list to be processed

• video_key -- the key name of field that stores sample video list to be processed

• image_bytes_key -- the key name of field that stores sample image bytes list to be processed

• query_key -- the key name of field that stores sample queries

• response_key -- the key name of field that stores responses

• history_key -- the key name of field that stores history of queries and responses

• min_closed_interval -- whether the min_val of the specified filter range is a closed interval. It's True by default.

• max_closed_interval -- whether the max_val of the specified filter range is a closed interval. It's True by default.

• reversed_range -- whether to reverse the target range [min_val, max_val] to (-∞, min_val) or (max_val, +∞). It's False by default.

get_keep_boolean(val, min_val=None, max_val=None)

compute_stats_batched(samples, *args, **kwargs)

process_batched(samples)

compute_stats_single(sample, context=False)

Compute stats for the sample which is used as a metric to decide whether to filter this sample.

参数:

• sample -- input sample.

• context -- whether to store context information of intermediate vars in the sample temporarily.

返回:

sample with computed stats

process_single(sample)

For sample level, sample --> Boolean.

参数:

sample -- sample to decide whether to filter

返回:

true for keeping and false for filtering

run(dataset, *, exporter=None, tracer=None, reduce=True)

class data_juicer.ops.Mapper(*args, **kwargs)

基类：OP

__init__(*args, **kwargs)

Base class that conducts data editing.

参数:

• text_key -- the key name of field that stores sample texts to be processed.

• image_key -- the key name of field that stores sample image list to be processed

• audio_key -- the key name of field that stores sample audio list to be processed

• video_key -- the key name of field that stores sample video list to be processed

• image_bytes_key -- the key name of field that stores sample image bytes list to be processed

• query_key -- the key name of field that stores sample queries

• response_key -- the key name of field that stores responses

• history_key -- the key name of field that stores history of queries and responses

process_batched(samples, *args, **kwargs)

process_single(sample)

For sample level, sample --> sample

参数:

sample -- sample to process

返回:

processed sample

run(dataset, *, exporter=None, tracer=None)

class data_juicer.ops.Deduplicator(*args, **kwargs)

基类：OP

__init__(*args, **kwargs)

Base class that conducts deduplication.

参数:

• text_key -- the key name of field that stores sample texts to be processed

• image_key -- the key name of field that stores sample image list to be processed

• audio_key -- the key name of field that stores sample audio list to be processed

• video_key -- the key name of field that stores sample video list to be processed

• image_bytes_key -- the key name of field that stores sample image bytes list to be processed

• query_key -- the key name of field that stores sample queries

• response_key -- the key name of field that stores responses

• history_key -- the key name of field that stores history of queries and responses

compute_hash(sample)

Compute hash values for the sample.

参数:

sample -- input sample

返回:

sample with computed hash value.

process(dataset, show_num=0)

For doc-level, dataset --> dataset.

参数:

• dataset -- input dataset

• show_num -- number of traced samples used when tracer is open.

返回:

deduplicated dataset and the sampled duplicate pairs.

run(dataset, *, exporter=None, tracer=None, reduce=True)

class data_juicer.ops.Selector(*args, **kwargs)

基类：OP

__init__(*args, **kwargs)

Base class that conducts selection in dataset-level.

参数:

• text_key -- the key name of field that stores sample texts to be processed

• image_key -- the key name of field that stores sample image list to be processed

• audio_key -- the key name of field that stores sample audio list to be processed

• video_key -- the key name of field that stores sample video list to be processed

• image_bytes_key -- the key name of field that stores sample image bytes list to be processed

• query_key -- the key name of field that stores sample queries

• response_key -- the key name of field that stores responses

• history_key -- the key name of field that stores history of queries and responses

process(dataset)

Dataset --> dataset.

参数:

dataset -- input dataset

返回:

selected dataset.

run(dataset, *, exporter=None, tracer=None)

class data_juicer.ops.Grouper(*args, **kwargs)

基类：OP

__init__(*args, **kwargs)

Base class that group samples.

参数:

• text_key -- the key name of field that stores sample texts to be processed

• image_key -- the key name of field that stores sample image list to be processed

• audio_key -- the key name of field that stores sample audio list to be processed

• video_key -- the key name of field that stores sample video list to be processed

• image_bytes_key -- the key name of field that stores sample image bytes list to be processed

• query_key -- the key name of field that stores sample queries

• response_key -- the key name of field that stores responses

• history_key -- the key name of field that stores history of queries and responses

process(dataset)

Dataset --> dataset.

参数:

dataset -- input dataset

返回:

dataset of batched samples.

run(dataset, *, exporter=None, tracer=None)

class data_juicer.ops.Aggregator(*args, **kwargs)

基类：OP

__init__(*args, **kwargs)

Base class that group samples.

参数:

• text_key -- the key name of field that stores sample texts to be processed

• image_key -- the key name of field that stores sample image list to be processed

• audio_key -- the key name of field that stores sample audio list to be processed

• video_key -- the key name of field that stores sample video list to be processed

• image_bytes_key -- the key name of field that stores sample image bytes list to be processed

• query_key -- the key name of field that stores sample queries

• response_key -- the key name of field that stores responses

• history_key -- the key name of field that stores history of queries and responses

process_single(sample)

For sample level, batched sample --> sample, the input must be the output of some Grouper OP.

参数:

sample -- batched sample to aggregate

返回:

aggregated sample

run(dataset, *, exporter=None, tracer=None)

class data_juicer.ops.Pipeline(*args, **kwargs)

基类：OP

Base class for Operators that represent a data processing pipeline.

__init__(*args, **kwargs)

Base class of operators.

参数:

• text_key -- the key name of field that stores sample texts to be processed.

• image_key -- the key name of field that stores sample image list to be processed

• audio_key -- the key name of field that stores sample audio list to be processed

• video_key -- the key name of field that stores sample video list to be processed

• image_bytes_key -- the key name of field that stores sample image bytes list to be processed

• query_key -- the key name of field that stores sample queries

• response_key -- the key name of field that stores responses

• history_key -- the key name of field that stores history of queries and responses

• index_key -- index the samples before process if not None

• batch_size -- the batch size for processing

run(dataset)

class data_juicer.ops.OPEnvSpec(pip_pkgs: List[str] | str | None = None, env_vars: Dict[str, str] | None = None, working_dir: str | None = None, backend: str = 'uv', extra_env_params: Dict | None = None, parsed_requirements: Dict[str, Requirement] | None = None)

基类：object

Specification of the environment dependencies for an operator.

__init__(pip_pkgs: List[str] | str | None = None, env_vars: Dict[str, str] | None = None, working_dir: str | None = None, backend: str = 'uv', extra_env_params: Dict | None = None, parsed_requirements: Dict[str, Requirement] | None = None)

Initialize an OPEnvSpec instance.

参数:

• pip_pkgs -- Pip packages to install, default is None. Could be a list or a str path to the requirement file

• env_vars -- Dictionary of environment variables, default is None

• working_dir -- Path to the working directory, default is None

• backend -- Package management backend, default is "uv". Should be one of ["pip", "uv"].

• extra_env_params -- Additional parameters dictionary passed to the ray runtime environment, default is None

• parsed_requirements -- a resolved version of requirements. It's a dict of req_name-resolved_info, where the parsed package info includes version/url/...

to_dict()

Convert the OPEnvSpec instance to a dictionary.

返回:

Dictionary representation of the OPEnvSpec instance

get_hash()

get_requirement_name_list()

data_juicer.ops.op_requirements_to_op_env_spec(op_name: str, requirements: List[str] | str | None = None, auto_recommended_requirements: List[str] | None = None) → OPEnvSpec

class data_juicer.ops.OPEnvManager(min_common_dep_num_to_combine: int | None = -1, conflict_resolve_strategy: ConflictResolveStrategy | str = ConflictResolveStrategy.SPLIT)

基类：object

OPEnvManager is a class that manages the environment dependencies for operators, including recording OP dependencies, resolving dependency conflicts, merging OP environments, and so on.

__init__(min_common_dep_num_to_combine: int | None = -1, conflict_resolve_strategy: ConflictResolveStrategy | str = ConflictResolveStrategy.SPLIT)

Initialize OPEnvManager instance.

参数:

• min_common_dep_num_to_combine -- The minimum number of common dependencies required to determine whether to merge two operation environment specifications. If set to -1, it means no combination of operation environments.

• conflict_resolve_strategy -- Strategy for resolving dependency conflicts, default is SPLIT strategy. SPLIT: Keep the two specs split when there is a conflict. OVERWRITE: Overwrite the existing dependency with one from the later OP. LATEST: Use the latest version of all specified dependency versions.

print_the_current_states()

Get the current states of OPEnvManager, including: - number of recorded OPs - number of used env specs - what OPs share the same env spec

返回:

A dictionary containing the current states of OPEnvManager

record_op_env_spec(op_name: str, op_env_spec: OPEnvSpec)

Record the OP environment specification for an operator.

参数:

• op_name -- Name of the operator

• op_env_spec -- OP environment specification

merge_op_env_specs(new_env_spec: OPEnvSpec)

Merge the OP environment specification for an operator with existing OP environment specification.

参数:

new_env_spec -- OP environment specification

can_combine_op_env_specs(first_env_spec: OPEnvSpec, second_env_spec: OPEnvSpec) → bool

Check if two OP environment specifications can be combined.

参数:

• first_env_spec -- Existing OP environment specification

• second_env_spec -- New OP environment specification

返回:

True if the two specifications can be combined, False otherwise

try_to_combine_op_env_specs(first_env_spec: OPEnvSpec, second_env_spec: OPEnvSpec)

Try to combine the OP environment specification for an operator with existing OP environment specification.

参数:

• first_env_spec -- Name of the operator

• second_env_spec -- OP environment specification

返回:

True if the two specifications can be combined, False otherwise

get_op_env_spec(op_name: str) → OPEnvSpec

Get the OP environment specification for an operator.

参数:

op_name -- Name of the operator

返回:

OP environment specification

data_juicer.ops.analyze_lazy_loaded_requirements(code_content: str) → List[str]

data_juicer.ops.analyze_lazy_loaded_requirements_for_code_file(code_file: str) → List[str]