Differences with torchtext.data.functional.custom_replace

torchtext.data.functional.custom_replace

torchtext.data.functional.custom_replace(replace_pattern)

For more information, see torchtext.data.functional.custom_replace.

mindspore.dataset.text.RegexReplace

class mindspore.dataset.text.RegexReplace(pattern, replace, replace_all=True)

For more information, see mindspore.dataset.text.RegexReplace.

Differences

PyTorch: Replace a part of string with given text according to regular expressions.

MindSpore: Replace a part of string with given text according to regular expressions.

Categories	Subcategories	PyTorch	MindSpore	Difference
Parameter	Parameter1	replace_pattern	-	Regex expression and string to replace stored in tuple
	Parameter2	-	pattern	Regex expression patterns
	Parameter3	-	replace	String to replace matched element.
	Parameter4	-	replace_all	If only replace first matched element

Code Example

list_a = ["Sentencepiece encode  aS  pieces"]

# PyTorch
from torchtext.data.functional import custom_replace

custom_replace_transform = custom_replace([(r'S', 's')])
print(list(custom_replace_transform(list_a)))
# Out: ['sentencepiece encode  as  pieces']

# MindSpore
import mindspore.dataset.text as text

transform = text.RegexReplace(pattern=r'S', replace='s')
print(transform(list_a))
# Out: ['sentencepiece encode  as  pieces']