File size: 6,671 Bytes
59ca01e 699f8e1 a254196 699f8e1 a254196 699f8e1 a254196 59ca01e a254196 59ca01e a254196 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 |
from copy import deepcopy
from typing import Dict, Iterable
from datasets import Dataset, DatasetDict, IterableDataset, IterableDatasetDict
from .generator_utils import ReusableGenerator
class Stream:
"""A class for handling streaming data in a customizable way.
This class provides methods for generating, caching, and manipulating streaming data.
Attributes:
generator (function): A generator function for streaming data.
gen_kwargs (dict, optional): A dictionary of keyword arguments for the generator function.
streaming (bool): Whether the data is streaming or not.
caching (bool): Whether the data is cached or not.
"""
def __init__(self, generator, gen_kwargs=None, streaming=True, caching=False):
"""Initializes the Stream with the provided parameters.
Args:
generator (function): A generator function for streaming data.
gen_kwargs (dict, optional): A dictionary of keyword arguments for the generator function. Defaults to None.
streaming (bool, optional): Whether the data is streaming or not. Defaults to True.
caching (bool, optional): Whether the data is cached or not. Defaults to False.
"""
self.generator = generator
self.gen_kwargs = gen_kwargs if gen_kwargs is not None else {}
self.streaming = streaming
self.caching = caching
def _get_initator(self):
"""Private method to get the correct initiator based on the streaming and caching attributes.
Returns:
function: The correct initiator function.
"""
if self.streaming:
if self.caching:
return IterableDataset.from_generator
else:
return ReusableGenerator
else:
if self.caching:
return Dataset.from_generator
else:
raise ValueError("Cannot create non-streaming non-caching stream")
def _get_stream(self):
"""Private method to get the stream based on the initiator function.
Returns:
object: The stream object.
"""
return self._get_initator()(self.generator, gen_kwargs=self.gen_kwargs)
def set_caching(self, caching):
self.caching = caching
def set_streaming(self, streaming):
self.streaming = streaming
def __iter__(self):
return iter(self._get_stream())
def unwrap(self):
return self._get_stream()
def peak(self):
return next(iter(self))
def take(self, n):
for i, instance in enumerate(self):
if i >= n:
break
yield instance
def __repr__(self):
return f"{self.__class__.__name__}(generator={self.generator.__name__}, gen_kwargs={self.gen_kwargs}, streaming={self.streaming}, caching={self.caching})"
def is_stream(obj):
return isinstance(obj, IterableDataset) or isinstance(obj, Stream) or isinstance(obj, Dataset)
def iterable_starter(iterable):
return iter(deepcopy(iterable))
class MultiStream(dict):
"""A class for handling multiple streams of data in a dictionary-like format.
This class extends dict and its values should be instances of the Stream class.
Attributes:
data (dict): A dictionary of Stream objects.
"""
def __init__(self, data=None):
"""Initializes the MultiStream with the provided data.
Args:
data (dict, optional): A dictionary of Stream objects. Defaults to None.
Raises:
AssertionError: If the values are not instances of Stream or keys are not strings.
"""
for key, value in data.items():
isinstance(value, Stream), "MultiStream values must be Stream"
isinstance(key, str), "MultiStream keys must be strings"
super().__init__(data)
def get_generator(self, key):
"""Gets a generator for a specified key.
Args:
key (str): The key for the generator.
Yields:
object: The next value in the stream.
"""
yield from self[key]
def unwrap(self, cls):
return cls({key: value.unwrap() for key, value in self.items()})
def to_dataset(self) -> DatasetDict:
return DatasetDict(
{key: Dataset.from_generator(self.get_generator, gen_kwargs={"key": key}) for key in self.keys()}
)
def to_iterable_dataset(self) -> IterableDatasetDict:
return IterableDatasetDict(
{key: IterableDataset.from_generator(self.get_generator, gen_kwargs={"key": key}) for key in self.keys()}
)
def __setitem__(self, key, value):
assert isinstance(value, Stream), "StreamDict values must be Stream"
assert isinstance(key, str), "StreamDict keys must be strings"
super().__setitem__(key, value)
@classmethod
def from_generators(cls, generators: Dict[str, ReusableGenerator], streaming=True, caching=False):
"""Creates a MultiStream from a dictionary of ReusableGenerators.
Args:
generators (Dict[str, ReusableGenerator]): A dictionary of ReusableGenerators.
streaming (bool, optional): Whether the data should be streaming or not. Defaults to True.
caching (bool, optional): Whether the data should be cached or not. Defaults to False.
Returns:
MultiStream: A MultiStream object.
"""
assert all(isinstance(v, ReusableGenerator) for v in generators.values())
return cls(
{
key: Stream(
generator.get_generator(),
gen_kwargs=generator.get_gen_kwargs(),
streaming=streaming,
caching=caching,
)
for key, generator in generators.items()
}
)
@classmethod
def from_iterables(cls, iterables: Dict[str, Iterable], streaming=True, caching=False):
"""Creates a MultiStream from a dictionary of iterables.
Args:
iterables (Dict[str, Iterable]): A dictionary of iterables.
streaming (bool, optional): Whether the data should be streaming or not. Defaults to True.
caching (bool, optional): Whether the data should be cached or not. Defaults to False.
Returns:
MultiStream: A MultiStream object.
"""
return cls(
{
key: Stream(iterable_starter, gen_kwargs={"iterable": iterable}, streaming=streaming, caching=caching)
for key, iterable in iterables.items()
}
)
|