CNTK/Source/Readers/Kaldi2Reader/DOCUMENTATION.txt

57 строки
1.7 KiB
Plaintext

Everything is the same as HTKMLFReader_linux, except:
********** Features **********
The features section is different:
features=[
dim=
rx=
scpFile=
featureTransform=
]
rx is a text file which contains:
one Kaldi feature rxspecifier readable by RandomAccessBaseFloatMatrixReader.
'ark:' specifiers don't work; only 'scp:' specifiers work.
scpFile is a text file generated by running:
feat-to-len FEATURE_RXSPECIFIER_FROM_ABOVE ark,t:- > TEXT_FILE_NAME
scpFile should contain one line per utterance.
If you want to run with fewer utterances, just shorten this file.
(It will load the feature rxspecifier but ignore utterances not present in scpFile).
featureTransform is the name of a Kaldi feature transform file:
Kaldi feature transform files are used for stacking / applying transforms to features.
An empty string (if permitted by the config file reader?) or the special string: NO_FEATURE_TRANSFORM
says to ignore this option.
********** Labels **********
The labels section is also different.
labels=[
mlfFile=
labelDim=
labelMappingFile=
]
Only difference is mlfFile. mlfFile is a different format now. It is a text file which contains:
one Kaldi label rxspecifier readable by Kaldi's copy-post binary.
********** Performance **********
# If you have 100000s of utterances or if your data is beyond 50 GB, you will need:
randomize=4320000
You don't need to do anything with UseAllDataForPreComputedNode (it's ok to use all data).
Read language ID DNN (stacked) train set 41728000 frames (130 GB) from scratch-raid in 597 seconds
Read language ID DNN (stacked) valid set 4350199 frames from scratch-raid in 106 seconds