o "j[1@s|ddlZddlmZddlmZGdddZGdddeZGd d d eZdd d Z GdddeZ GdddeZ dS)N)core)randpermc@s"eZdZdZdddZddZdS)Samplera8 An abstract class to encapsulate methods and behaviors of samplers. All sampler used by :code:`paddle.io.BatchSampler` should be a subclass of :code:`paddle.io.Sampler`, BatchSampler subclasses should implement following methods: :code:`__iter__`: return sample index iterably, which iterate over indices of dataset elements :code:`__len__`: the number of sample in :attr:`data_source` Args: data_source(Dataset, optional): this could be an instance of :code:`paddle.io.Dataset` other Python object which implemented :code:`__len__` for Sampler to get indices as the range of :attr:`dataset` length. Default None. Returns: Sampler: an iterable object for sample indices iterating Examples: .. code-block:: python >>> from paddle.io import Dataset, Sampler >>> class RandomDataset(Dataset): ... def __init__(self, num_samples): ... self.num_samples = num_samples ... ... def __getitem__(self, idx): ... image = np.random.random([784]).astype('float32') ... label = np.random.randint(0, 9, (1, )).astype('int64') ... return image, label ... ... def __len__(self): ... return self.num_samples ... >>> class MySampler(Sampler): ... def __init__(self, data_source): ... self.data_source = data_source ... ... def __iter__(self): ... return iter(range(len(self.data_source))) ... ... def __len__(self): ... return len(self.data_source) ... >>> sampler = MySampler(data_source=RandomDataset(100)) >>> for index in sampler: ... print(index) 0 1 2 ... 99 see `paddle.io.BatchSampler` see `paddle.io.DataLoader` NcC ||_dSN data_sourceselfr r ]/var/www/html/Deteccion_Ine/venv/lib/python3.10/site-packages/paddle/io/dataloader/sampler.py__init__W zSampler.__init__cCstr)NotImplementedErrorr r r r __iter__ZszSampler.__iter__r)__name__ __module__ __qualname____doc__rrr r r r rs A rc@s(eZdZdZddZddZddZdS) SequenceSamplera Iterate samples sequentially, yield :code:`0, 1, 2, ..., len(data_source) -1` generally, Args: data_source(Dataset): dataset to sample, this could be an instance of :code:`paddle.io.Dataset` other Python object which implemented :code:`__len__`. Returns: Sampler: a Sampler yield sample index sequentially Examples: .. code-block:: python >>> from paddle.io import Dataset, SequenceSampler >>> class RandomDataset(Dataset): ... def __init__(self, num_samples): ... self.num_samples = num_samples ... ... def __getitem__(self, idx): ... image = np.random.random([784]).astype('float32') ... label = np.random.randint(0, 9, (1, )).astype('int64') ... return image, label ... ... def __len__(self): ... return self.num_samples ... >>> sampler = SequenceSampler(data_source=RandomDataset(100)) >>> for index in sampler: ... print(index) 0 1 2 ... 99 see `paddle.io.Sampler` cCrrrr r r r rrzSequenceSampler.__init__cCsttt|jSr)iterrangelenr rr r r rszSequenceSampler.__iter__cC t|jSr)rr rr r r __len__rzSequenceSampler.__len__Nrrrrrrrr r r r ras + rc@s8eZdZdZ d ddZeddZdd Zd d ZdS) RandomSamplera Iterate samples randomly, yield shuffled indices, if :attr:`replacement=False`, yield shuffled indices of the whole data souce, if :attr:`replacement=True`, :attr:`num_samples` can set to specify the sample number to draw. Args: data_source(Dataset): dataset to sample, this could be an instance of :ref:`api_paddle_io_Dataset` or :ref:`api_paddle_io_IterableDataset` or other Python object which implemented :code:`__len__` to get indices as the range of :code:`dataset` length. Default None. replacement(bool, optional): If False, sample the whole dataset, If True, set :attr:`num_samples` for how many samples to draw. Default False. num_samples(int, optional): set sample number to draw if :attr:`replacement` is True, then it will take samples according to the number you set. Default None, disabled. generator(Generator, optional): specify a generator to sample the :code:`data_source`. Default None, disabled. Returns: RandomSampler: a Sampler yield sample index randomly. Examples: .. code-block:: python >>> import numpy as np >>> from paddle.io import Dataset, RandomSampler >>> np.random.seed(2023) >>> class RandomDataset(Dataset): ... def __init__(self, num_samples): ... self.num_samples = num_samples ... ... def __getitem__(self, idx): ... image = np.random.random([784]).astype('float32') ... label = np.random.randint(0, 9, (1, )).astype('int64') ... return image, label ... ... def __len__(self): ... return self.num_samples ... >>> sampler = RandomSampler(data_source=RandomDataset(100)) >>> for index in sampler: ... print(index) 56 12 68 ... 87 FNcCst||_||_||_||_t|jtstd|j|jdur%|s%tdt|jt r0|jdkr8td|jdS)Nz:expect boolean value for replacement, but got replacement=z>num_samples should not be specified while replacement is Falserz>num_samples should be a positive integer, but got num_samples=) r replacement _num_samples generator isinstancebool TypeError ValueError num_samplesint)r r rr&r!r r r rs* zRandomSampler.__init__cCs|jdur t|jS|jSr)r rr rr r r r&s  zRandomSampler.num_samplesc cst|j}|jr)t|jD]}zt|j}Wn ty"YdSw|VqdS|jrCtj j t ||jdd D]}|Vq;dStj j t ||dd D]}|VqQdS)NT)replaceF) rr r!rr&next StopIterationrnprandomchoiceZarangetolist)r niindexr r r rs4   zRandomSampler.__iter__cCs|jSr)r&rr r r rszRandomSampler.__len__)FNN) rrrrrpropertyr&rrr r r r rs2   rTcCsJt|tjr |}t|ttfrt|}t|tjs Jdt |j dks+Jd| d|j df}t |dks@Jdt |tjkrLJdt |tjkrXJdtj|dkd d }t |d kslJd |syt ||ksyJd ||jd d }g}t|j d D]}tj|j d ||||}||qt|S)Nz=weights should be paddle.Tensor, numpy.ndarray, list or tuplez$weights should be a 1-D or 2-D arraygz weights should be positive valuezweights shoule not be INFzweights shoule not be NaN)Zaxisrz#weights should have positive valueszUweights positive value number should not less than num_samples when replacement=False)r"rZ LoDTensornumpylisttupler+arrayZndarrayrshapereshapeallanyinfnansumrr,r-append)weightsr&rZ non_zerosZretsr0retr r r _weighted_samples8    rDc@s*eZdZdZd ddZddZddZd S) WeightedRandomSampleraG Random sample with given weights (probabilities), sampe index will be in range [0, len(weights) - 1], if :attr:`replacement` is True, index can be sampled multiple times. Args: weights(numpy.ndarray|paddle.Tensor|list|tuple): sequence of weights, should be numpy array, paddle.Tensor, list or tuple num_samples(int): set sample number to draw from sampler. replacement(bool): Whether to draw sample with replacements, default True Returns: Sampler: a Sampler yield sample index randomly by given weights Examples: .. code-block:: python >>> import numpy as np >>> from paddle.io import WeightedRandomSampler >>> np.random.seed(2023) >>> sampler = WeightedRandomSampler( ... weights=[0.1, 0.3, 0.5, 0.7, 0.2], ... num_samples=5, ... replacement=True ... ) >>> for index in sampler: ... print(index) 2 4 3 1 1 TcCsBt|tr |dkr tdt|tstd||_||_||_dS)Nrz(num_samples should be a positive integerz%replacement should be a boolean value)r"r'r%r#rBr&r)r rBr&rr r r rFs  zWeightedRandomSampler.__init__cCs$t|j|j|j}t|dSNr4)rDrBr&rrr;r.)r Zidxsr r r rOs zWeightedRandomSampler.__iter__cCs$t|jj|jjd}|j|SrF)r+prodrBr:r&)r mulr r r rUs zWeightedRandomSampler.__len__NTrr r r r rE!s  $ rEc@s.eZdZdZddZddZdefddZd S) SubsetRandomSamplera Randomly sample elements from a given list of indices, without replacement. Args: indices (sequence): a sequence of indices Examples: .. code-block:: python >>> import paddle >>> from paddle.io import SubsetRandomSampler >>> paddle.seed(2023) >>> sampler = SubsetRandomSampler(indices=[1, 3, 5, 7, 9]) >>> for index in sampler: ... print(index) 9 3 7 5 1 cCst|dkr td||_dS)NrzHThe length of `indices` in SubsetRandomSampler should be greater than 0.)rr%indices)r rKr r r rus  zSubsetRandomSampler.__init__ccs&tt|jD]}|j|VqdSr)rrrK)r r0r r r r|szSubsetRandomSampler.__iter__returncCrr)rrKrr r r rrzSubsetRandomSampler.__len__N)rrrrrrr'rr r r r rJZs rJrI) r6r+Z frameworkrZtensorrrrrrDrErJr r r r s  L6 j 9