espnet2.samplers.abs_sampler.AbsSampler

source

class espnet2.samplers.abs_sampler.AbsSampler(data_source: Sized | None = None)

Bases: Sampler, ABC

Abstract base class for defining custom samplers in PyTorch.

The AbsSampler class serves as a template for creating samplers that define how data should be sampled from a dataset. This class inherits from PyTorch’s Sampler class and mandates the implementation of the __len__ and __iter__ methods. Custom samplers must extend this class and provide specific sampling logic.

None

• Parameters:None
• Returns: None
• Yields: None
• Raises:
  • NotImplementedError – If __len__ or __iter__ methods are not
  • implemented in a subclass. –

####### Examples

To create a custom sampler, subclass AbsSampler and implement the required methods:

``

`

python class CustomSampler(AbsSampler):

def __len__(self) -> int: : return 100 # Example length

def __iter__(self) -> Iterator[Tuple[str, …]]: : for i in range(self._len_()): : yield (str(i),) # Example yield

sampler = CustomSampler() print(len(sampler)) # Output: 100 for item in sampler:

print(item) # Output: (‘0’,), (‘1’,), …, (‘99’,)

``

`

NOTE

The generate method returns a list of items generated by iterating over the sampler. It can be used for obtaining a deterministic sample based on a specific seed.

generate(seed)

Generates a list of samples from the sampler.

This method returns a list of samples generated by the sampler instance. The output will depend on the specific implementation of the sampler. The method takes a seed for random number generation to ensure reproducibility of the sampling process.

• Parameters:
  • seed (int) – A seed for the random number generator to ensure reproducible
  • results.
• Returns: A list of samples generated by the sampler.
• Return type: List[Tuple[str, …]]

####### Examples

>>> sampler = SomeConcreteSampler()
>>> samples = sampler.generate(seed=42)
>>> print(samples)
[('sample1',), ('sample2',), ('sample3',)]

NOTE

The behavior of this method is dependent on the specific implementation of the AbsSampler subclass.