Hi everyone,
I was working on the tutorial last night using an4 dataset, SphinxTrain and the S3 Flat decoder.
I basically ran through everything and ended up with 100% sentence Error at the end. When I looked at the .match file, it looked like I got null for everything.
I think i must have missed some step in the process. I'm going to go over quickly what I did, and hopefully someone can spot the missing part. My plan tonight is to keep reading the various documentation, esp the tutorial to see if i missed anything, and perhaps try out rm1 as well.
I have a dir called tutorial that contains am4, s3flat, and SphinxTrain. I configured and made both s3flat and SphinxTrain according to the instructions.
I ran the setup_tutorial script from the s3flat dir (perl scripts/setup_tutorial.pl an4)
Then i ran the make_feats script (perl scripts_pl/make_feats.pl -ctl etc/an4_train.fileids) and the RunAll script. The RunAll script seemed to run without any errors (I watched the logfile.)
Then I ran make_feats but for the test data.
and finally ran decode, which is where I saw that it didn't match anything.
I would appreciate any help or pointers to other relevant threads that I may have missed. Thanks,
Rachel
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I found that troubleshooting page that explains error messages and warnings. I've seen this warning a few times in 04.cd_schmm_untied/an4.1-1.bw.log:
WARNING: "mod_inv.c", line 368: Model inventory n_density not set; setting to value in mixw file, 1.
Troubleshooting Help excerpt:
Message
WARNING: "../mod_inv.c", line 308: # of tied states in mdef file, 755 != # of mixing weight sets, 2965, in ../hmms/09.cd_chmm/1d/mixw
WARNING: "../mod_inv.c", line 328: Model inventory n_density not set; setting to value in mixw file, 1.
Explanation You are either using the wrong model definition file for this part of the training, or the wrong model parameter files. check these out.
So it says i am using either the wrong model def file or the wrong model param files. ANy clues on where to look?
Thanks
Rachel
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
So this morning i fiddled with running the scripts one at a time and stared at the log files. I also ran some c executables as well (ones in script #02).
Somehow this worked, even though I don't really know what I did differently...
Anyway, it was great to see a sentence error of only 65% (vs 100%).
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I wasn't able to reply early and I am sorry for this.
If you are using an4 test set, many people were able to repeat that in their own machines with few problems. So I am guessing there must be something wrong related to the settings.
Would you like to send us the log in step 2 and 3? In that way, we would be able to help you to debug.
-a
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi everyone,
I was working on the tutorial last night using an4 dataset, SphinxTrain and the S3 Flat decoder.
I basically ran through everything and ended up with 100% sentence Error at the end. When I looked at the .match file, it looked like I got null for everything.
I think i must have missed some step in the process. I'm going to go over quickly what I did, and hopefully someone can spot the missing part. My plan tonight is to keep reading the various documentation, esp the tutorial to see if i missed anything, and perhaps try out rm1 as well.
I have a dir called tutorial that contains am4, s3flat, and SphinxTrain. I configured and made both s3flat and SphinxTrain according to the instructions.
I ran the setup_tutorial script from the s3flat dir (perl scripts/setup_tutorial.pl an4)
Then i ran the make_feats script (perl scripts_pl/make_feats.pl -ctl etc/an4_train.fileids) and the RunAll script. The RunAll script seemed to run without any errors (I watched the logfile.)
Then I ran make_feats but for the test data.
and finally ran decode, which is where I saw that it didn't match anything.
I would appreciate any help or pointers to other relevant threads that I may have missed. Thanks,
Rachel
Let's do this,
Look at the log in each step and see whether you could find something wrong first. Chances are you might be able to figure out the issue yourself.
When you come to us again, send us the log you feel there were problems. We will help you to diagnose.
-gj
I found that troubleshooting page that explains error messages and warnings. I've seen this warning a few times in 04.cd_schmm_untied/an4.1-1.bw.log:
WARNING: "mod_inv.c", line 368: Model inventory n_density not set; setting to value in mixw file, 1.
Troubleshooting Help excerpt:
Message
WARNING: "../mod_inv.c", line 308: # of tied states in mdef file, 755 != # of mixing weight sets, 2965, in ../hmms/09.cd_chmm/1d/mixw
WARNING: "../mod_inv.c", line 328: Model inventory n_density not set; setting to value in mixw file, 1.
Explanation You are either using the wrong model definition file for this part of the training, or the wrong model parameter files. check these out.
So it says i am using either the wrong model def file or the wrong model param files. ANy clues on where to look?
Thanks
Rachel
So this morning i fiddled with running the scripts one at a time and stared at the log files. I also ran some c executables as well (ones in script #02).
Somehow this worked, even though I don't really know what I did differently...
Anyway, it was great to see a sentence error of only 65% (vs 100%).
Hi Rachel,
I wasn't able to reply early and I am sorry for this.
If you are using an4 test set, many people were able to repeat that in their own machines with few problems. So I am guessing there must be something wrong related to the settings.
Would you like to send us the log in step 2 and 3? In that way, we would be able to help you to debug.
-a