VoxForge
Hello,
Increased the value to 896 and tried both codes but still no luck. Can you tell me how much horsepower you've got on the comp where you ran the file successfully?
Thanks!
Fabien
--- (Edited on 7/14/2011 9:46 am [GMT-0500] by Visitor) ---
Hi Fabien,
The computer I ran this file on has a Core 2 Duo 2.0 GHz processor with a 2 GB RAM.
I am hence puzzled why it's not working on your computer. Let's go through the details one by one.
First, I have modified the code in the repository, I would like you to checkout out a fresh copy of it.
Second let's just have this one test in your batch file, i.e. remove oov_numbers.wav and the other tests from your tests for now.
Now could you run your test and provide me with the logs?
-
Apurv
--- (Edited on 7/14/2011 2:42 pm [GMT-0500] by Visitor) ---
Hello Apurv,
OK, so I've downloaded the new code, copied the lib, edited the classpath file, test ran OK with the numbers text/wav files. I then proceeded to remove said tests from the folders and batch and used call_of_the_wild_chapter_01 instead.
Here is the log:
http://pastebin.com/Aju44EHm
I'm not a Java programmer. However I'm a bit confused that the logs read "Total Memory= 253440" vs. "-Xmx896m" in my Eclipse ini file.
I hope it helps. Thanks!
Fabien
--- (Edited on 7/15/2011 8:26 pm [GMT-0500] by Visitor) ---
Hi Fabien,
It seems the memory logs are actually a bit confusing after all !
Total Memory ideally should report the total memory currently being used by your JVM. -Xmx on the other hand is supposed to be a cap on the memory that JVM _can_ use.
Your logs suggest that somehow the JVM is using only a small part of the memory available. But I recently found out that these logs can be deceptive( http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=4686462 ).
I am still trying recreate your error on my computer. Will get back to you as soon as I have something.
-
Apurv
--- (Edited on 7/16/2011 2:50 pm [GMT-0500] by apurvtwr) ---
Hi Fabien,
As I indicated in my last mail, your JVM does not have enough memory available though you are allowing that in eclipse.ini.
Please try the following, and check if it works. If not provide me with the logs again.
In your eclipse project, under run configurations> Arguments> VM arguments specify -Xmx896m
What you faced seems to be a common problem and the above should fix it.
-
Apurv
--- (Edited on 7/17/2011 12:55 pm [GMT-0500] by apurvtwr) ---
Hello Apurv,
Thank you for your help. Turns out using the run configuration options was without effect also. I ended up editing the Ant file.
I've added a couple of variables:
<!-- Setting Memory Options -->
<property name="xms" value="-Xms512m"/>
<property name="xmx" value="-Xmx512m"/>
And updated the "run" part like so:
<target name="run" depends="compile">
<java jar="${bin_dir}/aligner.jar" fork="true">
<jvmarg value="${xms}"/>
<jvmarg value="${xmx}"/>
</java>
</target>
Memory logs now read what you would expect and the program ran successfully! :) It also cut my running time down to just 11 minutes. Thank you for the program! I might be using it intensely in the next few months, I'll keep you posted if you're interested.
Quick question: are there plans to dump the results to a text file (as opposed to the console) and/or to support a different file format? (something like .lrc or sub files).
Sincerely,
Fabien
--- (Edited on 7/18/2011 3:40 pm [GMT-0500] by Visitor) ---
Hi Fabien,
Your adjustments in Ant build file seem correct.You may let us know if you face any problem while using this piece of code.
The run time earlier was large because most of it went for GC task due to lack of memory. 11 minutes is quite the time it should take in the current state.
Dumping the result in a file should not be a problem, you can make those minor adjustments in LongAudioAligner.java
Currently we are focussing on reducing the runtime memory requirements. You can follow the development and use a more efficient code once it is available in the repository.
Regards,
Apurv
--- (Edited on 7/19/2011 4:28 am [GMT-0500] by apurvtwr) ---