Forum Replies Created
-
AuthorPosts
-
Hola Pilar!
I ended up copying my own version of sptk 3.7 and world. Than in the config file wrote a path to wherever i copied them to and it worked. I attached both in the zip file. Have a look maybe it will work for you as well.
Sorry, I checked it again and I see that DNNGEN only creates cmps and next produces this error. Maybe it is supposed to create all necessary acoustic features bap, f0 and mfcc as well? where do I specify what the output should be? Is it in a duration config file by any chance?
In addition, the file parameter_generation.py has (at least in my version – Merlin for Ossian) in the main() function following paths:
in_file_list = [‘/afs/inf.ed.ac.uk/group/project/dnn_tts/mtl_dnn/gen/dnn_2500_601_229/hvd_678.cmp’]
and a variable var_file_dict a dictionary which values are some other paths.
I believe they should be updated? If yes, could someone explain what is their function?
Anyone assist, please?
thanks,
Norbertthanks Simon,
all sorted I had to add the path to WORLD as well.
I was misled by /this/path/does/not/exist/ I thought it was a msg from the command line and not the actual path, ah sure happens.Fair play to you,
much appreciated!Training run_dnn_bottleneck.py we encountered several problems which we managed to solve and run the training. However, the process remaining a little bit vague, hece a few questions and comments that we would like to address.
First, we commented out the import of the following modules: ms_dnn, ms_dnn_gv, sdae, as we are going to use just DNN model.
Second, we had to change the tpye of learning rate to numpy.asarray with type float 32. Could you please explain why it is necessary?
Third, we had to uncomment lines 574, 576, 577. Those lines are responsible for creating label files in the dnn_streams/binary_label_number and dnn_streams/nn_no_silence_lab_number which are essential for the bottleneck training. We would like to know what nn_no_silence labels are and why are they required for the bottleneck training while they are not required for DNN training?
Fourth, while training we get a WARNING: no silence found! Has it anything to do with question 3?
Fifth, when specifying the size of hidden layers in config file, we have changed the size of the 5th of the 6 hidden layers to 512. The other 5 remain 1024. Is there any rule of thumb when deciding on the size of the bottleneck layer or is it established empirically?thanks for your reply,
Norbert / GiorgiaHi,
Sorry, didn’t realise it, as I have never used svn before.
Decided to create my own copy then: svn co https://svn.ecdf.ed.ac.uk/repo/inf/dnn_tts.
When trying to run the file I get a msg that our old good friend bandmat is
not installed – will try to sort it now…
As for svn upgrade / update dont know what I am doing wrong but it doesn’t work for me.hi Srikanth,
thanks for your reply which triggered another question, namely svn update.
I typed svn help in the command line which provides among many options update or up
argument, so I tried the below but get an error msg.ppls-atlab-015:ss_dnn_letter_based s1574060$ svn update /Volumes/Network/courses/ss/dnn/dnn_tts/
svn: E155021: This client is too old to work with the working copy at
‘/Volumes/Network/courses/ss/dnn/dnn_tts’ (format 31).
You need to get a newer Subversion client. For more details, see
http://subversion.apache.org/faq.html#working-copy-format-changeCould you tell me what the correct syntax for the update is, please?
Sorry, I meant the letters in each utterance are separated by a white space.
I also changed all non ascii characters, so HTK does not complain.there was a white space after dictionary_seq\
I am getting good at it!thanks for the tip!
model A emission probs: 0.18,0.12,0.12,0.2,0.2,0.2
model B emission probs: 0.24,0.05, 0.09, 0.1,0.1,0.18Cecilia/Simon,
your x values should be the observations, so 5 4 2 3 3 3,
then you have given most likely state sequence for each model.
Hence, I took each observation and calculated emission prob using mean and sd of the most likely model in the sequence. However, not all my results exactly match the answers. I blame the calculator :).
Simon, was my way of doing it right? I am confused, since some calculations are like in your answer and some different? Are we going to talk about it tomorrow?Thanks for your help. The problem was in the text editor I used. It was adding characters that was not recognised by HTK.
-
AuthorPosts