train.ExpertChunkedSampler Class Reference

Inheritance diagram for train.ExpertChunkedSampler:

Collaboration diagram for train.ExpertChunkedSampler:

Public Member Functions
	__init__ (self, dataset_indices, chunk_size=1_000_000)
	__iter__ (self)
	__len__ (self)

Public Attributes
	num_samples = len(dataset_indices)
	chunk_size = chunk_size
	num_chunks = math.ceil(self.num_samples / self.chunk_size)
list	spatial_chunks = []

Detailed Description

Spatially sort expert indices, then randomize at chunk/local levels.

Constructor & Destructor Documentation

__init__()

train.ExpertChunkedSampler.__init__	(		self,
			dataset_indices,
			chunk_size = 1_000_000 )

Member Function Documentation

__iter__()

train.ExpertChunkedSampler.__iter__

(

self

)

__len__()

train.ExpertChunkedSampler.__len__

(

self

)

Member Data Documentation

chunk_size

train.ExpertChunkedSampler.chunk_size = chunk_size

num_chunks

train.ExpertChunkedSampler.num_chunks = math.ceil(self.num_samples / self.chunk_size)

num_samples

train.ExpertChunkedSampler.num_samples = len(dataset_indices)

spatial_chunks

list train.ExpertChunkedSampler.spatial_chunks = []

---

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

• training/train.py