I have a sendump file created with the sphinxtrain which is the largest file of my acoustic model (3.3mo) .
To reduce the size of this file, I 'd like to build a 8bit senone dump files.
There 's a section in the Sphinx User Guide but it's not clear in my mind..
Where is the file pdf32to8 ?
Can I run directly pdf32to8 on my sendump file created during the training ?
When I cluster down to 8bits my 32bits senone dump file, is the size reduce by 4 ? (3.3 mo 32bit ->? 0.825 mo 8bit )
Thx for your help...sorry for my poor froggy english
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
My understanding is that the documentation is old and that the sendump file created by the current SphinxTrain is the 8bit dump already. If you follow the directions to create an 8bit dump file, the first thing it does is destroy the sendump file which you got from sphinxtrain. So, use what you get directly from sphinxtrain..
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Anonymous
-
2001-09-12
Ok, thank you for this information but are you sure of that ?
'cos I 'm trying to reduce the size of my acoustic model and that was the only way I found to do this.
Do you know other way to reduce the size of the acoustic model ?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I have a sendump file created with the sphinxtrain which is the largest file of my acoustic model (3.3mo) .
To reduce the size of this file, I 'd like to build a 8bit senone dump files.
There 's a section in the Sphinx User Guide but it's not clear in my mind..
Where is the file pdf32to8 ?
Can I run directly pdf32to8 on my sendump file created during the training ?
When I cluster down to 8bits my 32bits senone dump file, is the size reduce by 4 ? (3.3 mo 32bit ->? 0.825 mo 8bit )
Thx for your help...sorry for my poor froggy english
My understanding is that the documentation is old and that the sendump file created by the current SphinxTrain is the 8bit dump already. If you follow the directions to create an 8bit dump file, the first thing it does is destroy the sendump file which you got from sphinxtrain. So, use what you get directly from sphinxtrain..
Ok, thank you for this information but are you sure of that ?
'cos I 'm trying to reduce the size of my acoustic model and that was the only way I found to do this.
Do you know other way to reduce the size of the acoustic model ?