fastreid.data.samplers.data_sampler.TrainingSampler Class Reference

Inheritance diagram for fastreid.data.samplers.data_sampler.TrainingSampler:

Public Member Functions
	__init__ (self, int size, bool shuffle=True, Optional[int] seed=None)
	__iter__ (self)

Protected Member Functions
	_infinite_indices (self)

Protected Attributes
	_size
	_shuffle
	_seed
	_rank
	_world_size

Detailed Description

In training, we only care about the "infinite stream" of training data.
So this sampler produces an infinite stream of indices and
all workers cooperate to correctly shuffle the indices and sample different indices.
The samplers in each worker effectively produces `indices[worker_id::num_workers]`
where `indices` is an infinite stream of indices consisting of
`shuffle(range(size)) + shuffle(range(size)) + ...` (if shuffle is True)
or `range(size) + range(size) + ...` (if shuffle is False)

Definition at line 15 of file data_sampler.py.

Constructor & Destructor Documentation

__init__()

fastreid.data.samplers.data_sampler.TrainingSampler.__init__	(		self,
		int	size,
		bool	shuffle = True,
		Optional[int]	seed = None )

Args:
    size (int): the total number of data of the underlying dataset to sample from
    shuffle (bool): whether to shuffle the indices or not
    seed (int): the initial seed of the shuffle. Must be the same
        across all workers. If None, will use a random seed shared
        among workers (require synchronization among all workers).

Definition at line 26 of file data_sampler.py.

    def __init__(self, size: int, shuffle: bool = True, seed: Optional[int] = None):
        """
        Args:
            size (int): the total number of data of the underlying dataset to sample from
            shuffle (bool): whether to shuffle the indices or not
            seed (int): the initial seed of the shuffle. Must be the same
                across all workers. If None, will use a random seed shared
                among workers (require synchronization among all workers).
        """
        self._size = size
        assert size > 0
        self._shuffle = shuffle
        if seed is None:
            seed = comm.shared_random_seed()
        self._seed = int(seed)
 
        self._rank = comm.get_rank()
        self._world_size = comm.get_world_size()
 

Member Function Documentation

__iter__()

fastreid.data.samplers.data_sampler.TrainingSampler.__iter__

(

self

)

Definition at line 45 of file data_sampler.py.

    def __iter__(self):
        start = self._rank
        yield from itertools.islice(self._infinite_indices(), start, None, self._world_size)
 

_infinite_indices()

fastreid.data.samplers.data_sampler.TrainingSampler._infinite_indices ( self )

protected

Definition at line 49 of file data_sampler.py.

    def _infinite_indices(self):
        np.random.seed(self._seed)
        while True:
            if self._shuffle:
                yield from np.random.permutation(self._size)
            else:
                yield from np.arange(self._size)
 
 

Member Data Documentation

_rank

fastreid.data.samplers.data_sampler.TrainingSampler._rank

protected

Definition at line 42 of file data_sampler.py.

_seed

fastreid.data.samplers.data_sampler.TrainingSampler._seed

protected

Definition at line 40 of file data_sampler.py.

_shuffle

fastreid.data.samplers.data_sampler.TrainingSampler._shuffle

protected

Definition at line 37 of file data_sampler.py.

_size

fastreid.data.samplers.data_sampler.TrainingSampler._size

protected

Definition at line 35 of file data_sampler.py.

_world_size

fastreid.data.samplers.data_sampler.TrainingSampler._world_size

protected

Definition at line 43 of file data_sampler.py.

---

The documentation for this class was generated from the following file:

• smreid/fastreid/data/samplers/data_sampler.py