Speech:Spring 2015 Mohamed Fadlalla Log


 * Home
 * Semesters
 * Spring 2015
 * Proposal
 * Report
 * Information - General Project Information
 * Experiments - List of speech experiments

Week Ending February 3, 2015
Read more logs. Get a better understanding of the system. Run the first experiment.
 * Task:

Read most of the logs. I think I spent a good amount of time reading logs this week, and hopefully start playing with the system next week. I successfully ssh'd into caesar. I also spent some time reading about linux nfs and RAID. Knowing how the machines in our network communicate will be very helpful.
 * Results:

Work with the system group to install Redhat in one of the Dell PowerEdge 1750 machines. It seems like the installation process is documented now and ready to be used!
 * Plan:

The speach system doesn't have a testing environment. Also, all the configuration managed "manually".
 * Concerns:

Week Ending February 10, 2015
Run train on the new Caesar and the old Caesar and compare the results.
 * Task:


 * Results:

I was getting this error "awk: cmd. line:1: fatal: cannot open file `etc/002.dic' for reading (No such file or directory)". After skimming through the CreateTranscript.pl script, I figured it was looking for the data corpus inside /mnt/main/corpus/switchboard/. Well, the the data corpus I specified doesn't live inside /mnt/main/corpus/switchboard any more, it's now in /mnt/main/corpus/switchboard/old. So I had to pick different data corpus. I also needed to edit the PruneDictionary4.pl script because it wasn't pointing to the correct dictionary location. It was looking for it at /mnt/main/corpus/dist/custom when it exists at /mnt/main/corpus/switchboard/dist/dict/custom. Seems like it's finally training! Finished running train on Caesar and Asterix through Caesar. I also updated the wiki page to reflect the change I made to the pruneDictionary4.pl script.
 * 02/06/15: Unable to run a train on Caesar. Following this tutorial http://foss.unh.edu/projects/index.php/Speech:Run_Train_Setup_Scrip
 * 02/07/15: Still unable to run a train. Added the to the prepareExperiment3 script to the scripts page.


 * Plan:

Install Redhat in Methusalix Installing Redhat on Methusulix is taking more that it should. Forest suggested changing the DVD drive. Melissa spent almost two hours to get working but she couldn't. I think the whole group should focus on this issue for next week.
 * Concerns:

Week Ending February 17, 2015
Run a decode and get the result of the 0260 experiment.
 * Task:

Due to power outage in the server room we couldn't recover the network setting in Verleihnix
 * Results:

Unmount Brutus from Caesar. Mount Verleihnix to Brutus. Verleihnix needs to be connected to the Internet in order to activate redhat. This is probably is going to take some time!
 * Plan:
 * Concerns:

Week Ending February 24, 2015
Decode the 0260 experiment and get the result. Create a network bridge on Caesar in order for Verleihnix to access the Internet.
 * Task:


 * Results:
 * 02/19/15: Spend 90 minutes in the server room trying to connect Verleihnix to the Internet. I am still not convinced this is a Redhat vs Opensuse issue. I think there is something in Caesar's network setting that we don't understand. I will keep working on this over the weekend.
 * 02/23/15: Verleihnix is connected to the Internet!! Finally I figured out the problem was the kernel ip routing table. Setting Caesar as our default gateway in the eth0 config file wasn't just enough, we needed to add it to the kernel ip routing table.
 * 02/24/15: Decoded the 0260/001 experiment in Caesar. This will hellp us to compare the systems performance.

Decode the 0260/001 experiment. Start copying data to brutus. None.
 * Plan:
 * Concerns:

Week Ending March 3, 2015

 * Task:


 * Results:


 * 02/25/2015: Spent 3 hours after the class to figure out how to connect Brutus to the Internet. After configuring Brutus' network cards and plunging the red cable, the server was still not getting Internet packets. The problem is solved by help from Professor Jonas. We let DHCP pick up an IP for eth0 and configure it, connected eth0 to the Internet, turned off DHCP, and this worked.
 * 03/02/2015: We just figure out that the /home folder has more space than / . We spent another three hours trying to re-partition Brutus so the /mnt/main would have more disk space. No luck! We will keep working on this tomorrow.
 * 03/02/2015: Met with Melissa to start the copying process. Chris figured out how to partition Brutus.


 * Plan:

I am concerned that we won't be able to finish on time. Unexpected things keep happening and we have two days left.
 * Concerns:

Week Ending March 10, 2015

 * Task:

Can't mount Verleihnix on Brutus because Brutus is not activated. Can't Run a train on Brutus. Waiting for Sam to fix training script. Can't Run a train on Verleihnix through Brutus. Waiting for Sam to fix the training script. Recopy the Exp(0260-0267) directories Nfs mount Verleihnix on Brutus Do two trains on Brutus and Verleihnix
 * Results:
 * 03/05/15: Can't activate Brutus. Wrong keys or permissions. Need to consult with IT department.
 * 03/09/15: Melissa successful activated Brutus. However, we are still unable to mount Verleihnix. I spend some time in school to solve this problem. I will keep working on it tomorrow and hopefully will be able to fix it by Wednesday.
 * 03/10/15: Solved the nfs configuration problem.
 * Plan:

Meet with the group tomorrow before class to do the switch over. There are minor steps that need to be completed before switching Caesar to Brutus.
 * Concerns:
 * 03/05/15: A little bit concerned. Look at the task section!

Week Ending March 24, 2015

 * Task:


 * Results:


 * Plan:


 * Concerns:

Week Ending March 31, 2015
Run 5hr train in the new Caesar. Run another 5hr in Verleihnix. Meet with the systems group to make a plan for mounting the remaining drones on Caesar. Run 125hr train.
 * Task:

Keep working on the 125hr trains None
 * Results:
 * 03/29/15: Running 5hr train in the new Caesar. After finishing this train I will run another one on Verleihnix. I am also doing some reading about speech in the sphinx page. http://cmusphinx.sourceforge.net/wiki/tutorialconcepts
 * 03/30/15: Nfs mounted the rest of the drones, except for Obelix, on the new Caesar. Also worked with the systems group to create the user accounts on these drones. Rafik had an issue installing emacs on Obelix so we checked Caesar to make sure it's activated. Caesar is activated but we still get the "not active" message when using yum.
 * Plan:
 * Concerns:

Week Ending April 7, 2015
Run 125hr train.
 * Task:
 * Results:
 * 04/05/13: As the semester is coming to an end I need to make sure that we are following the plan we had to finish Redhat installation. It seems like we are doing good till now, but a couple of things left. I spent time reading over our documentation and try to figure out what is left to be done. I will meet with the systems team next time to make a new plan for finalizing Redat switchover.
 * 04/06/13: Having problem getting the result of 5hr train (segmentation fault). I usually see this kind of issues in C programs, but I don't know what's happening here. I have to spent more time debugging this. I am also unable to run a new train. It complains about file locations (file doesn't exist) error. I looked int the path the file is there!
 * 04/06/13: In the process of running a new train
 * Plan:


 * Concerns:

Week Ending April 14, 2015
Copy the old Corpa data to Brutus.
 * Task:


 * Results:
 * 04/11/15: Started coping the old directory from Brutus to Caesar. I wanted to do this from home over the weekend, but it didn't work exactly as I planned to. I couldn't keep the SSH session open to copy the whole data from Brutus to Caesar. To solve this problem I executed the job in the background. There are all kind of wrong could go with this. There are many jobs running on Caesar right now. A low priority task coming from Brutus will even make it slower. After awhile I checked Caesar, but I couldn't find all files.
 * 04/11/15: Spent time to solve slow logging problem on Caesar. When Professor Jonas and I first connected Caesar to the Internet, we turned DHCP on. This was needed because Caesar couldn't recognize the switch's IP address. When turning DHCP on, some how, Caesar sent a packet to the outside world and when the packet came back, it got all the information needed about the network. This typically includes the IP of the switch. However, turning DHCP on also resulted in a change to the name servers on /etc/resolv.conf. After copying over the original ones from Brutus (the old Caesar) and restarting the network service, Caesar now is faster.
 * 04/11/15: Started copying the old Corpa data from Brutus to Caesar. This can take awhile, so I am not going to wait for it. I also worked with Melissa to activate Obelix and install gcc on it.

Run 125hr train. None
 * Plan:
 * Concerns:

Week Ending April 21, 2015
Keep running trains
 * Task:

Keep training and decoding none
 * Results:
 * 04/19/15: Running 125hr train on idefix
 * 04/20/15: Checking in
 * 04/21/15: Decoding
 * Plan:
 * Concerns:

Week Ending April 28, 2015

 * Task:


 * Results:
 * 04/26/15: Checking in
 * 04/27/15: Unable to access Caesar via cisunix. When trying to ssh into cisunix I get unable to connect error message. Doing a quick port scanning, seems like tcp port 22 is closed.
 * 04/28/15: Doing research about the options we have for energy optimization. In the classroom, professor Jonas spoke about how much energy the drones consume. The amount of energy all the machines consume together, could be reduce if we figure out an efficient way to allocate our resources. He purposed that machines to be used as they needed. All of the unneeded machines should be powered off until someone (or some software) requests them.
 * Plan:


 * Concerns:

Week Ending May 5, 2015
Consult with the Celtics team to see if any tasks need to be completed. Go over the system page again to see if any information need to be added to the page.
 * Task:


 * Results:


 * Plan:
 * 5/4/15: Reading logs
 * Concerns: