Speech:Spring 2017 Huong Ha Log

From Openitware
Jump to: navigation, search


Week Ending February 7th, 2017

Task

2/4

Read everyones logs to see where everyone is at. Look back at previous semesters to see what we have as a general basis for Torque.

2/5

Read Logs and write a log.

2/6

Read Log and write a log. See where the status of the draft of proposal is at. Communicate with group members.

2/7

Look at Summer 2013 links for Torque information.

Results

2/4

It looks like a lot of people are currently familiarizing themselves with Caesar and previous logs to get a better understanding of Speech. There is a lot of information on Torque in the Misc. section of Summer 2013 including what is working, what is not working, what were their plans, and resources. Using torque for parellelzation was a task in the Summer of 2015, but was never implemented. It did not seem to be a big task for other semesters, because there is no documentation of it in other semesters.

2/5

Read logs and everyone seems to still be doing research and getting used to the environment. Wrote a log.

2/6

Read logs. Spoke to Sharayah today to see where our draft proposal was and what was the status on it and it still needs some updates on the timeline and could use more background detail. We decide our group goal by the end of class next week is to all successfully run a train.

2/7

The first link for under Torque updates for the Torque documentation no longer brings you to an running page. The second link has lots of information on torque including how to install/configure it, submitting and managing jobs, managing nodes, integrating schedulers, and accounting records. From what I can see from Tommy's logs, his what's next section for torque updates were never accomplished.

Plans

2/4

Get a better understanding and general basis of LDA Read documents in the resource area of Summer 2013 which include links to Torque documentation, logs, and a walkthrough of a job. Get familiar with environment.

2/5

Continue to do more research on the where Torque is currently at and what has been done to it in the pass. Continue to do research on LDA.

2/6

Update proposal. Continue to do research on Torque and LDA. Attempt to run a train.

2/7

Same as before.

Concerns

2/4

Using an unfamiliar environment and inaccurate information due to lack of consistent documentation on Torque.

2/5

Same as before.

2/6

Successfully running a train and same as before.

2/7

Same as before.

Week Ending February 14, 2017

Task

2/11

See if G++ is backwards compatible with GCC. Look back on previous logs to see what has been done on GCC and on what machine. Look to see if there was any effort on trying to install G++. Get a better understanding of GCC and G++.

2/12

Look more into the pros and cons of GCC and G++. See if G++ can be downloaded as a tar file. Read logs.

2/13

Research GCC, G++, and Clang to see how they compare to each other then produce a brief results table to show findings.

2/14

Read logs and compare gcc and g++ directories.

Results

2/11

GCC is installed on majestix as of last semester. Going back even further there seems to be no evidence of G++ ever being installed or even the consideration.

GCC can only compile C programs and C++ with a certain flag, and G++ compiles C and C++ programs. From what I have research so far it seems like G++ is backwards compatible with GCC, because when G++ is compiling programs it does call the GCC core and enables automatic linking with the C++ library. However, GCC can link the C++ library on its own commanding a flag -lstdc++. Both can be stand alone compilers though.

2/12

G++ can be downloaded as tar file following this link at the bottom of the page~> http://www.ccp14.ac.uk/ccp14admin/gnu/gcc.html. The installation of a C/C++ more will be beneficial to the Modeling groups effort in trying to implement the Recurrent Neural Network Language Models (RNNL) toolkit considering it is written in C/C++ scripts.

From what my research has shown so far, G++ is preferable for compiling C++ programs. I also have notice while looking the comparison between G++ and GCC that there are a lot more comparisons between GCC and Clang. There does not seem to be a lot of research done on performance comparison, but I will continue to look.

From reading the logs, I noticed that Greg said that the tools group will look into the necessary actions for implementation of RNNL. Currently, the tools group were instructed to look into G++ only. I wonder if Greg was thinking of any other tools his group needs to aid the implementation of RNNL.

2/13

Drivers Basic Info Performance Size Newest version
GCC *Installed on Majestix
*Unknown version (Not documented)
*GCC 5.4 is faster than G++ 5.4 according to https://benchmarksgame.alioth.debian.org/u64q/c.html Installed size is 41KB GCC 6.3 is the newest version
G++ Installed size is 34KB G++ 5.4 is the newest version

2/14

GCC installed would include 12 directories:

/usr/bin/c89-gcc
/usr/bin/c99-gcc
/usr/bin/gcc
/usr/bin/gcov
/usr/bin/x86_64-linux-gnu-gcc
/usr/share/doc/cpp/README.Bugs
/usr/share/doc/gcc
/usr/share/man/man1/c89-gcc.1.gz
/usr/share/man/man1/c99-gcc.1.gz
/usr/share/man/man1/gcc.1.gz
/usr/share/man/man1/gcov.1.gz
/usr/share/man/man1/x86_64-linux-gnu-gcc.1.gz

Can be found here ~>http://packages.ubuntu.com/precise/amd64/gcc/filelist

G++ installed would include 3 directories:

/usr/bin/g++
/usr/bin/x86_64-linux-gnu-g++
/usr/share/doc/g++

Can be found here ~>https://packages.debian.org/wheezy/amd64/g++/filelist

It seems like installing G++, but not take up a lot of space so the decision kind of easier to make especially since we are really only putting three directories and taking up 34KB on which machine or drone we are given.

Plan

2/11

Continue to research more about GCC and G++ and get a pros and cons list going See if G++ can be downloaded in a Tar ball.

2/12

Try to come up with a data table of G++, GCC, and Clang including current version necessary libraries, performance, and a general information about each compiler.

2/13

Download G++and give it a test trial and to see how many directories it has. Research and continue to update my results table. Research Pocket Sphinx.

2/14

Give up on trying to research Clang, because the size was in the thousands for KB. Have a more concrete and detailed table of information on G++ and GCC.

Concerns

2/11

None at the moment.

2/12

Finding credible information on the performance of these compilers.

2/13

Same as before.

2/14

None at the moment.

Week Ending February 21, 2017

Task

2/16

Finish writing report on G++ and GCC to show the benefits of installing G++ and show current state of Majestix. I read Sharayah's log and apparently Jonas would like us to look into what is currently installed on Obelix. So I will also be looking back on previous semesters to see what is the current state of Obelix.

2/19

Read logs and write a log. Continue to update the GCC and G++ comparison report.

2/20

Read logs and get a general understanding of pocket sphinx since I feel like not focused at all on it.

2/21 Checking in and reading logs.

Results

2/16

From the group meeting on Wednesday, we have combined our research on GCC and G++. We are currently in the process of writing a report that compares the two, gives basic information on them, and helps us give a recommendation on whether we should go through with an install of G++ or not. With our current research we are very much leaning towards downloading G++. If we get Obelix instead of Majestix, I'm leaning towards just downloading G++ only and not GCC. From what I read of RNNLM here ->http://www.fit.vutbr.cz/~imikolov/rnnlm/rnnlm-demo.pdf, it only needs a c++ compiler.

2/19

Read Sharayah's logs and it seems that she is getting a lot of updates on the machines. I need to talk to her to see what is going on with that. I am also having trouble logging on to Majestix with my username and password. There is a lot of progress on the comparison report such as comparing performance, newest versions, and the size and the directories both offer. We should have a recommendation followed with the finished version of the report to Jonas on Wednesday.

2/20

Pocket Sphinx is recommended handheld and mobile devices, because it is very light weight. However, it also works just as well on a desktop. Pocket Sphinx is written in C and according to ~>http://cmusphinx.sourceforge.net/wiki/tutorialpocketsphinx GCC is one of its necessary dependencies. Although, I think if we went down the path of only installing G++ and leaving out GCC it would not cause a problem. There are other dependencies such as automake, autoconf, libtool, bison, and swig. Dependencies can also be flagged out during installation if necessary. I think the only dependencies that I see us keeping are GCC and the python development package. When installing Pocket Sphinx, Sphinx base needs to be installed as well and they both have to be the same version.

2/21

I am still unable to ssh in Majestix. Although, it may just be my internet, because I also cannot ssh into caesar. And I know that we now have access to it because Sharayah logged onto majestix. I will try to do it tomorrow on a different network.

Plan

2/16

To finish the GCC and G++ report so we are able to back up our suggestions and show Jonas our findings. Look back on previous semesters to see what is the current state of Obelix.

2/19

Have a final version of the GCC and G++ comparison report. Successfully log onto Majestix. Research more on Pocket Sphinx.

2/20

Still not able to log onto Majestix with my account so hopefully that can be done by the end of Wednesday. Add finishing touches on GCC and G++ comparison report.

2/21

Successfully log onto Majestix. See if there is anything else I can add to the GCC and G++ comparison report before Jonas takes a look at it.

Concerns

2/16

At first I was relatively concerned about not getting Majestix, but I think Obelix maybe be a better opportunity since it has no c++ compiler on there. So if we install GCC or G++, we can just install one and not worry about having two installs taking up memory and cpu.

2/19

My only concern is not having a machine that the tools group can log onto successfully by Wednesday.

2/20

Not being able to log onto Majestix and Rome in a timely manner. Installing software onto the machines kind of makes me nervous, but looking back on previous semesters and it doesn't seem that bad.

2/21

Not being able to ssh into Majestix at my house. Pretty much same as before.

Week Ending February 28, 2017

Task

2/23

Copy files from mnt/main over to usr/local on Majestix. Unmount Majestix from Caesar.

2/26

Run a train on Majestix for before G++ installation comparison.

2/27

Run Language model to complete my train. Read logs and update progress.

2/28

Finish creating the language model so I can finish running this train. May have to start a new experiment.

Results

2/23

There were a few problems trying to copy to over the files from mnt/main to usr/local. Mostly, because we did not know what to do at first. We needed to be logged in as root on Caesar before we tried to copy over the files, which we did not do at first. Then we recursively copied the files from mnt/main to Majestix using the command, scp -r /mnt/main/* majestix:/local/usr.

Once we copied the files over we tried to unmount Majestix from Caesar. This is where more major problems occurred. Once I put in the command unmount -a, I got error saying that the device was busy. Of course, I immediately assumed that Majestix was still mounted to Caesar and so did professor Jonas when I showed him the error. And apparently, this error displayed on every machine when you try to unmount it. Jonas later listed the directories after he did the umount -a, there was nothing. Then he mounted it back on Caesar and it showed a bunch of directories.

So unmount -a does work even though it does say the device is busy after sending the command unmount -a command. To make sure it is definitely unmounted just list the directories by using the command ~> dir /mnt/main after unmount and make sure that there are no directories and vice versa for mounting it back onto Caesar.

2/26

Currently in the process of trying to run a train on Majestix before we try to install G++ on there. I will update this with results when it is successful.

2/27

I was having some trouble running the addExp script, but so did Sharayah. Apparently the experiment group is working on that so I just manually put my experiment onto wiki and the master experiment log. I was did the first part of running a training successfully in the sub experiment 005 and will update when I finish the language model. Looks like Jeff is having a problem still unmounting majestix, will have to check in with him on that and make sure he is able to. Jeremy quite possible has found a way to installs G++ without internet so we will be looking into that soon.

2/28

I am running into an error while trying to create my language model in sub experiment 005. When I try coping the script the create the language, cp -i /mnt/main/scripts/user/lm_create.pl. I get this error ~>

cp: missing destination file operand after `mnt/main/scripts/user/lm_create.pl' Try `cp --help' for more information.

I do not really know why it is doing that. I have googled the error and it has not lead me to a correct answer. There are two area to finds instructions on creating the language model. It's quite possible that mixing the old instructions with new ones may have ruined the copy for me. This is why I may just try and run another experiment.

Plan

2/23

Attempt to run a train on Majestix to capture it before the G++ install.

2/26

Finish running a train on Majestix and talk to Jeff about trying to install G++ on there.

2/27

Run a complete train so the tools group can start trying to install G++ on majestix next wednesday.

2/28

Have a train run on Majestix by the start of class tomorrow.

Concerns

2/23

Not being able to successfully run a train.

2/26

This train not being a success.

2/27

Running into errors while trying to complete the train.

2/28

Not being able to run a succesful train on majestix by the start of class tomorrow.

Week Ending March 7, 2017

Task

3/2

Run a train and decode on majestix before installing G++.

3/5

Reading logs and checking in.

3/6

Run a train and decode on Obelix for comparison purposes.

Results

3/2

Sharayah and I both ran into a problem when trying to run our decode on majestix. We were had a problem running this command ~>nohup run_decode.pl . Apparently it was not able to find a certain library, but that was the came from a larger problem. And the larger problem was copying over more of Caesar than just mnt/main/local. This caused the script to get confused when trying to locate the library, because there were many duplicate directories. However, Jonas was able to fix it by moving our duplicates into a delete directory.

We were successful in running the decode for our train on majestix, but we ran the decode on caesar because at that moment Jonas had not helped us resolve our duplicate directories problem.

Even though we were successful the results are now kind of useless, because majestix is unable to install programs due to not being to access yum installs or apt-get. So now we are going to get obelix instead which may involve us installing GCC and G++ on it.

3/5

Read logs and checked in.

3/6

It seems like there are no files that we can find that needs to be copied over so we gave the systems group the okay to wipe it clean and we will be starting anew with Obelix.

Once the train and decode is done. I will update this.

Plan

3/2

Try and locate the text files on majestix that we need to move over to obelix. Try and run a train on obelix.

3/5

Same as before.

3/6

Hopefully get GCC and G++ installed on Obelix on Wednesday.

Concerns

3/2

Not being able to locate the file we need to copy over if there even are any.

3/5

Same as before.

3/6

None at the moment

Week Ending March 21, 2017

Task

3/19

Checking in

3/20

Reading logs and checking in

Results

3/19

Checking in

3/20

My laptop is currently broken so I am writing this from my phone. Sharayah has run a train on Obelix for the after GCC installation, but did not do it for untrained data. The documentation for the GCC installation is done. So we will be moving on with an installation of G++ on Obleix on Wednesday.

Plan

3/19

Checking in

3/20

Attempt to run a train on untrained data and get my laptop fixed so I can do so.

Concerns

3/19

Checking in

3/20

None at the moment except for my laptop.

Week Ending March 28, 2017

Task

3/22

The goal for today is to install g++.

3/26

Check in and read logs.

3/27

Check in and read logs.

3/28

Run a train on unseen data on Obelix.

Results

3/22

At first apt-get was not recognized in the command line, but the reason was Obelix was not connected to the internet. However, we continued to run into issues with it not recognizing the command. So we just reinstalled gcc using "yum install gcc" and then tried to install g++ again using "yum install gcc-c++". After, we checked to see if it installed properly by checking the version of gcc and g++ with this command "gcc --version" and "g++ --version".

3/26

Checked in and read logs.

3/27

Checked in and read logs.

3/28

The connection to the server keeps getting broken, making very hard to run a train in general never mind a train on unseen data. Also, nohup continues to generate a nohup.out file along with this print out "nohup: ignoring input and appending output to `nohup.out'". However, if I look at what is running in the background that that command is running. I am going to try and get this finished by class tomorrow and add more to this results.

Plan

3/22 ' Run a train for after installation performance check and run another train, but for untrained data.

3/26

Same as before.

3/27

Check to see if my group needs any installs on Asterix or Miraculous to be done. Run a train on unseen data since Sharayah has already ran a train on seen data on Obelix. Talk to Jonas about installing G++ on Caesar.

3/28

Same as before and maybe get a start on gathering information on pocketsphinx.

Concerns

3/22

I do not have many concerns at the moment except for the trains not running successfully. But lately they have been running successfully so this is just a mild concern.

3/26

None at the moment.

3/27

There has yet to be a train that ran or ran successfully on unseen data in the tools group so far so my concern is it not running successfully.

3/28

Not being able to successfully run a train on unseen data on Obelix

Week Ending April 4, 2017

Task

3/30

Do install of minianaconda on asterix and install numpy and scipy libraries.

4/2

Check in and read logs.

4/3

Write a proposal for the installation of G++ on Caesar.

4/4

Email Jonas about the G++ install on Caesar and figure out a different way of getting miniconda on Miraculix so that every user not just root has access to it.

Results

3/30

Jeff and I installed minianaconda on Asterix in the /usr/bin/ logged in as root with Caesar unmounted referring to Vitali's logs. We ended up using the link he used which is, https://www.atlantic.net/community/howto/install-python-2-7-centos-anaconda/.

We essentially just followed the instructions on there, but tweaked a few things. For example, we used the command "wget" on a miniconda package from https://repo.continuum.io/miniconda/Miniconda-latest-Linux-x86_64.sh, but instead of the 64bit we did a wget on the 32bit. So the command we used looked like this, wget https://repo.continuum.io/miniconda/Miniconda2-latest-Linux-x86.sh.

The second part had us run the shell script we just did a wget on. We did not do the command that was shown on the website instead we changed the permissions on it and made the script an executable with the command, chmod +x Miniconda2-latest-Linux-x86.sh. Then we ran it using, ./Miniconda2-latest-Linux-x86.sh and it will execute and finish with agreeing to a license agreement that you have to say yes to.

The 3rd part is setting up where we want it installed. The prompted location is a miniconda directory in root so we changed it to /root/usr/bin. However, that brings up the problem of on being able to access miniconda as root, because I logged into my own account on Asterix and typed in python and was only able to access python 2.6.6 instead of 2.7 which comes with the miniconda. Now we are going to be fixing that with Vitali.

4/2

Currently people are having a problem sshing into Asterix and I just noticed that it is also an issue on Miraculix. This also puts Jeff's concerns about our installation of miniconda on Asterix being the instigator, to rest because it is an issue on both machines and we did not touch Miraculix. However, when I try to log onto Asterix it just hangs and when I try to ssh into Miraculix I get an actual error which is, ssh: connect to host miraculix port 22: No route to host.

4/3

The snapshots for the G++ installation has been in the tools group log so we are using that information to write up a proposal for Jonas's approval for the installation of G++ on Caesar. The proposal currently has the number of files and the name of those files that the installation impacts. I recently asked Andrew if the systems group is able to turn Brutus into Caesar if anything goes awry with the installation. Once I get the conformation from the systems group the proposal will go to Jonas.

4/4

I emailed Jonas to take a look at our proposal for the installation for G++ on Caesar so hopefully he got to look at it and we can get a confirmation tomorrow. The install should not take that long so it would be ideal to do the install at the beginning of the class. I was thinking of other ways to install miniconda on Miraculix and Vitali had mention a way using WinScp which is essentially a file transfer application. I think it may be a cleaner approach than what we did before so we can just download the miniconda tar files onto our local machine, transfer them through WinScp, and then unpack them on Miraculix. Vitali already installed the numpy and scipy libraries on Asterix so we no longer need to worry about that.

Plan

3/30

Get the miniconda access problem fixed and then install numpy and scipy. Create a proposal for the installation for g++ on Caesar and get in touch with the systems group to see if they are able to turn Brutus into Caesar for worst case scenario of our install.

4/2

Write up G++ installation on Caesar proposal and get it into Jonas ASAP so the tools group can do a quick install on Caesar at the beginning of class Wednesday. We also still need to get the miniconda path issue fixed and the installs for numpy and scipy should be quick.

4/3

Try to find a solution to the miniconda access problem. Get G++ installation proposal to Jonas and install ASAP. Run a train on Miraculix for a before comparison before Jeffrey and I try to install miniconda on it.

4/4

Getting the path fixed on Asterix is not really a major issue, but it would be ideal for non root users to access it as well. However, I think we will back burner this issue since it is not a major issue at the moment. The major things that Jeff and I have to take care is the G++ installation on Caesar and the installation of miniconda on Miraculix while tracking it.

Concerns

3/30

Getting this miniconda issue fixed in a timely manner.

4/2

Same as before. Also, I am concerned about not being able to ssh into Asterix and Miraculix, because at the moment the empire team are just sitting ducks without our machines.

4/3

My concerns involve fixing the path of the miniconda installation. It is not ideal for only root to have access to it. There probably is a way to fix. I just don't want to mess up anything else in the process of fixing it.

4/4

The only concern I have is my installation plan for miniconda on Miraculix does not go as smoothly as I thought it would so hopefully it does.

Week Ending April 11, 2017

Task

4/6

Start URC poster and install miniconda on Miraculix and track it.

4/8

Take snapshots of Miraculix to track the directories before the miniconda install. Install miniconda on Miraculix.

4/10

Download and move miniconda 64bit over to Miraculix

4/11

Checking in

Results

4/6

The tools group currently has a good draft of the URC poster, but I think do not think we are finished yet. Jeff and I also tried to the get the install of miniconda on Miraculix. So at first I download miniconda and transferred the file onto my user folder on Caesar using Filezilla. Then we tried to copy it over to Miraculix and unpack it there in the /usr/local/miniconda directory, but we continued to get an error when trying to copy it over. The error was saying that it could not find the destination that I was trying to copy it over to.

4/8

I took the snapshots of Miraculix and they are in the snapshots directory in the root directory Miraculix. I was able to copy over the miniconda executable from my user directory on Caesar with the scp command. I made a directory called miniconda and the executable was copied into there. I continuously ran into an error while trying to install it. It was always give me an error saying that the directory or file already existed whenever I tried to run this command, sh Miniconda2-latest-Linux-x86.sh -b -p /usr/local/miniconda. I found out later that the error was due to the fact that executing this file automatically creates a miniconda directory to contain the install and I had already created a miniconda directory so these two were conflicting each other. So I resolved that problem by moving the executable out of the miniconda folder I created into /usr/local and then ran the command, sh Miniconda2-latest-Linux-x86.sh -b -p /usr/local/miniconda and it worked fine... up until the point where I got another error.

This error was about not being able to import a certain library and I knew that this had to do with the warning that came at the begining of the install saying that this is a 64bit device and I was trying to install the 32bit version of miniconda, which is the same version that we successfully installed on Asterix. So now I know that Miraculix needs the 64bit version instead.

4/10

I downloaded the miniconda 64bit and move it onto Miraculix like I did before. Jeff is going to try and do the install tomorrow.

4/11

Checking in

Plan

4/6

Copy over the miniconda package onto Miraculix in the usr/local/miniconda directory and install it. Then install the numpy and scipy libraries on there as well. Take snapshots before and after install.

4/8

Get the 64bit version of miniconda onto Miraculix and install it.

4/10

Install miniconda onto Miraculix and install the numpy and scipy libraries. Take after installation snapshots.

4/11

Install GCC and G++ on Caesar

Concerns

4/6

Not figuring out a solution to this error, but it's a mild concern. I am pretty sure this is going to be resolved fairly quickly.

4/8

I am worried that this may not be the solution to the error, but I am fairly confident it is.

4/10

None at the moment.

4/11

None at the moment

Week Ending April 18, 2017

Task

4/13

Install miniconda and take after installation snapshots. Then install the scipy and numpy libraries.

'4/16

Checking in

4/17

Get pre-GCC snapshots

4/18

Checking in

Results

4/13

We were able to get the installation of miniconda on Miraculix done and get snapshots of it. After we successfully installed the scipy and numpy libraries. Jeff and I ended up installing miniconda in the wrong directory again, because it would continue to give the error of the file or directory already existing so it ended up being in /root/local/usr which does not allow you to access it if you are not root. Again this is not ideal, but for our purposes of just using it for LDA and everyone having access to root it is not that big of a deal, but not ideal.

The right way to install miniconda though is to have the executable in the /usr/local directory and then use this command while in that directory ~> bash Miniconda-latest-Linux-x86_64.sh(For the 64bit machines) or bash Miniconda2-latest-Linux-x86.sh. Say yes to the terms and then do not use the default path for the install instead use /usr/local/miniconda and it will automatially created a miniconda folder and put the install in there.

As for the scipy and numpy installation, it is a simple command ~> conda install <numpy or scipy>

To know that the install is done correctly with the right path, you should be able to login as your user account, type python and it should show you the version of python that you have, which should be python 3.

4/16

Checked in

4/17

It seems that we did not get snapshots of Obelix before the GCC install. Assuming that Obelix probably just looks like one of the other drones with /mnt/main copied over, Sharayah and I were able to take before snapshots on Miraculix and will be uploading them onto the wiki. There are snapshots from the year before of pre-GCC installation so we will compare to see if they are similar.

4/18

Checked in

Plan

4/13

See where we are at for the installation of GCC and G+. See if the systems group is prepared for that and then ask Jonas again if we can do so.

4/16

Same as before

4/17

Same as before and maybe do some more research on pocketsphinx

4/18

Same as before

Concerns

4/13

My only concern is the delay of our installation, because I would like to get it done ASAP. However, I have informed the modeling group that they should not wait on us for it, because it may take some time.

4/16

None at the moment

4/17

None at the moment

4/18

None at the moment

Week Ending April 25, 2017

Task

4/22

Checking in

4/23

Read logs and check in

4/24

See if Jonas has responded yet so we can proceed to do the installs on Caesar.

4/25

See where we are at with Jonas for the installation on GCC and G++ on Caesar

Results

4/22

Checking in

4/23

It seems like we have yet to have gotten a reply from Jonas yet. So we will have to continue waiting for him to give the okay.

4/24

It does not seem like Jonas has responded. Hopefully we will be able to get the okay to do it before or on Wednesday.

4/25

It seems like Jonas responded wanting to know which libraries it hit with in the list of directories we give him. Sharayah replied to him with that so hopefully we will be able to move forward with the installs tomorrow.

Plan

4/22

See where we are at for getting the okay to install GCC and G++

4/23

Hopefully install GCC and G++ on Caesar

4/24

Look back on our install documentations and see if they need any clarification or organization. And he same as before.

4/25

Install GCC and G++ on Caesar

Concerns

4/22

None at the moment

4/23

Same as before

4/24

Not getting the okay from Jonas is a major concern, because the tools team would not like to delay this any longer.

4/25

Same as before.

Week Ending May 2, 2017

Task

4/29

Checking in

4/30

Read logs and check in

5/1

Run a 5 hour train on Idefix

5/2

Run 5 hour train on Idefix and update results report once the modeling group settles on a result because it seems like they are still trying to get the best result that they can with the time we have left.

Results

4/29

Checked in

4/30

We have already ran a successful 5hour train on Caesar for comparison purposes so all we have left to do is complete a 5 hour train on Idefix. However, there was a 30hr train running on it and either it is still finishing up or there is currently another train running on it.

5/1

Greg has mainly been working and running train on Idefix and he is currently running another 30 hour train on it. Looks like we need to coordinate a time with him to be able to run our train on it.

5/2

Greg is still running 30 hour trains on Idefix. We'll try to coordinate a time with him so we can try and run our 5 hour train. The empire group is also running 30 hour trains on Asterix and talking with the data group to see if our word error rate can improve with the work they have done.

Plan

4/29

Run a 5 hour train on Idefix

4/30

Same as before and to update the results report

5/1

Run 5 hour train on Idefix and update results report once the modeling group settles on a result because it seems like they are still trying to get the best result that they can with the time we have left.

5/2

Same as before

Concerns

4/29

None at the moment

4/30

None at the moment

5/1

I am concerned about not having the time to be able to run the 5 hour train on Idefix, because Greg has been working on it exclusively and I would hate to interrupt him since he is also trying to get the best results he possible can on it.

5/2

Same as before

Week Ending May 9th, 2017

Task

5/6

Checking in

5/7

Read logs

5/8

Checking in

Results

5/6

Checking in

5/7

The empire team has run into errors with the new dictionary and the tools documentation looks like it is finished.

5/8

Checking in

Plan

5/6

Checking in

5/7

It seems like Idefix is free at the moment planning on trying to get a 5 hour train running on that tomorrow.

Concerns

5/6

Checking in

5/7

Hopefully the errors with the dictionary can be solved before Wednesday. However, I think that even if it were to get fixed before Wednesday our team would not have enough time to run another train for the final results.

5/8

Checking in