<html>
  <head>
    <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
  </head>
  <body text="#000000" bgcolor="#FFFFFF">
    <p>Hi,</p>
    <p>Because I managed to get enough of the corpus annotated on a
      smaller cluster to experiment with on the next part of the
      pipeline, I didn't run this on abel, but now I am planning to get
      the rest of the corpus annotated on abel, so I tried it out.  It
      seems to work so far, we need a minimum of 10GB to load all the
      embeddings but so far the nodes haven't choked on that
      requirement.  Thanks.<br>
    </p>
    <p>Yours,<br>
      --Asad.<br>
    </p>
    <br>
    <div class="moz-cite-prefix">On 2018-10-10 12:08 PM, Stephan Oepen
      wrote:<br>
    </div>
    <blockquote type="cite"
cite="mid:CA+_Fm6JUYY6vbtGi0oAQBJ05axHrqUCduz0Pp5G5-vnPnWH-3w@mail.gmail.com">
      <div dir="ltr">
        <div>hi again, asad,</div>
        <div><br>
        </div>
        <div>sorry, this took me a little while to return to!  and
          thanks again for your feedback; this is useful to us in the
          larger picture too :-)!</div>
        <div><br>
        </div>
        <div>our NLPL python modules are indeed created as virtual
          environments, but that should of course not prevent users from
          adding software using ‘pip install --user’.  i believe i have
          addressed that problem in the ‘nlpl-play’ module now; at
          least, i was able to install pyhocon on top of it, but into my
          home directory.</div>
        <div><br>
        </div>
        <div>that being said, i then added pyhocon and tensorflow_hub
          (which we should include by default, i now realize) to
          ‘nlpl-play’ ... so if you are still game, please give getting
          your SRL pipeline running on Abel another shot!</div>
        <div><br>
        </div>
        <div>best wishe, oe</div>
        <div><br>
        </div>
        <div><br>
        </div>
        <div><br>
        </div>
      </div>
      <br>
      <div class="gmail_quote">
        <div dir="ltr">On Thu, Oct 4, 2018 at 10:41 PM Asad Sayeed <<a
            href="mailto:asayeed@mbl.ca" moz-do-not-send="true">asayeed@mbl.ca</a>>
          wrote:<br>
        </div>
        <blockquote class="gmail_quote" style="margin:0 0 0
          .8ex;border-left:1px #ccc solid;padding-left:1ex">
          <div text="#000000" bgcolor="#FFFFFF">
            <p>Hi,</p>
            <p>Yes, typing python2 or python2.7 works now. Thanks!</p>
            <p>Unfortunately I cannot install tensorflow_hub because it
              thinks that I am in a virtualenv and I'm pretty sure I am
              not. But I need that too. And pyhocon. <br>
            </p>
            <p>The software I am trying to run is here, if you're
              wondering:</p>
            <p><a class="m_-1413699099110551646moz-txt-link-freetext"
                href="https://github.com/luheng/lsgn" target="_blank"
                moz-do-not-send="true">https://github.com/luheng/lsgn</a></p>
            <p>It needs the latest stuff.  But under python2.7.  But
              it's really very good...<br>
            </p>
            <p>Yours,<br>
              --Asad.<br>
            </p>
            <br>
            <div class="m_-1413699099110551646moz-cite-prefix">On
              2018-10-04 10:27 PM, Stephan Oepen wrote:<br>
            </div>
            <blockquote type="cite">
              <div>
                <div dir="auto">i think you are getting the wrong
                  ’python‘ binary; try ’python2‘ or ’python2.7‘.  use</div>
              </div>
              <div dir="auto"><br>
              </div>
              <div dir="auto">type -all python2</div>
              <div dir="auto"><br>
              </div>
              <div dir="auto">to confirm you end up with the binary from
                ’.../play/1.11/bin/‘.</div>
              <div dir="auto"><br>
              </div>
              <div dir="auto">oe</div>
              <div dir="auto"><br>
              </div>
              <div><br>
                <div class="gmail_quote">
                  <div dir="ltr">On Thu, 4 Oct 2018 at 22:23 Asad Sayeed
                    <<a href="mailto:asayeed@mbl.ca" target="_blank"
                      moz-do-not-send="true">asayeed@mbl.ca</a>>
                    wrote:<br>
                  </div>
                  <blockquote class="gmail_quote" style="margin:0 0 0
                    .8ex;border-left:1px #ccc solid;padding-left:1ex">Hi,<br>
                    <br>
                    I tried it, including the module purge, but then
                    when I start Python 2.7 <br>
                    and try to load tensorflow, I get:<br>
                    <br>
                    ***************<br>
                    <br>
                    -bash-4.1$ python<br>
                    Python 2.7.10 (default, Jul  1 2015, 11:02:23)<br>
                    [GCC Intel(R) C++ gcc 4.4 mode] on linux2<br>
                    Type "help", "copyright", "credits" or "license" for
                    more information.<br>
                     >>> import tensorflow as tf<br>
                    Traceback (most recent call last):<br>
                       File "<stdin>", line 1, in <module><br>
                       File <br>
"/projects/nlpl/software/play/1.11/lib/python2.7/site-packages/tensorflow/__init__.py",
                    <br>
                    line 22, in <module><br>
                         from tensorflow.python import
                    pywrap_tensorflow  # pylint: <br>
                    disable=unused-import<br>
                       File <br>
"/projects/nlpl/software/play/1.11/lib/python2.7/site-packages/tensorflow/python/__init__.py",
                    <br>
                    line 49, in <module><br>
                         from tensorflow.python import pywrap_tensorflow<br>
                       File <br>
"/projects/nlpl/software/play/1.11/lib/python2.7/site-packages/tensorflow/python/pywrap_tensorflow.py",
                    <br>
                    line 74, in <module><br>
                         raise ImportError(msg)<br>
                    ImportError: Traceback (most recent call last):<br>
                       File <br>
"/projects/nlpl/software/play/1.11/lib/python2.7/site-packages/tensorflow/python/pywrap_tensorflow.py",
                    <br>
                    line 58, in <module><br>
                         from
                    tensorflow.python.pywrap_tensorflow_internal import
                    *<br>
                       File <br>
"/projects/nlpl/software/play/1.11/lib/python2.7/site-packages/tensorflow/python/pywrap_tensorflow_internal.py",
                    <br>
                    line 28, in <module><br>
                         _pywrap_tensorflow_internal =
                    swig_import_helper()<br>
                       File <br>
"/projects/nlpl/software/play/1.11/lib/python2.7/site-packages/tensorflow/python/pywrap_tensorflow_internal.py",
                    <br>
                    line 24, in swig_import_helper<br>
                         _mod =
                    imp.load_module('_pywrap_tensorflow_internal', fp,
                    pathname, <br>
                    description)<br>
                    ImportError: <br>
/cluster/software/VERSIONS/python_packages-2.7_6/lib64/libc.so.6: <br>
                    version `GLIBC_2.14' not found (required by <br>
/projects/nlpl/software/play/1.11/lib/python2.7/site-packages/tensorflow/python/_pywrap_tensorflow_internal.so)<br>
                    <br>
                    <br>
                    Failed to load the native TensorFlow runtime.<br>
                    <br>
                    See <br>
                    <a
href="https://www.tensorflow.org/install/install_sources#common_installation_problems"
                      rel="noreferrer" target="_blank"
                      moz-do-not-send="true">https://www.tensorflow.org/install/install_sources#common_installation_problems</a><br>
                    <br>
                    for some common reasons and solutions.  Include the
                    entire stack trace<br>
                    above this error message when asking for help.<br>
                     >>><br>
                    <br>
                    ***************************<br>
                    <br>
                    So there is still a glibc discrepancy on the abel
                    login nodes?<br>
                    <br>
                    Thanks!<br>
                    <br>
                    Yours,<br>
                    --Asad.<br>
                    <br>
                    On 2018-10-04 01:29 PM, Stephan Oepen wrote:<br>
                    > hi asad,<br>
                    ><br>
                    > thanks for your feedback!  i am glad you like
                    the Abel cluster, even<br>
                    > though it cannot currently run what you would
                    like to do :-).<br>
                    ><br>
                    > it seems it did not take very long for an NLPL
                    user to question our<br>
                    > assumption that folks should be able to make do
                    with just Python 3.x<br>
                    > these days.  challenge accepted!<br>
                    ><br>
                    > i cannot yet promise this will play out in the
                    long run, nor do i want<br>
                    > to promise we will commit to supporting it. 
                    but it appears i may have<br>
                    > managed to apply the ‘glibc gymnastics’ to a
                    TensorFlow installation<br>
                    > that supports both Python 2.7 and 3.5.<br>
                    ><br>
                    > i would encourage you to try the following:<br>
                    ><br>
                    > module purge<br>
                    > module use -a
                    /projects/nlpl/software/modulefiles<br>
                    > module load nlpl-play/1.11/2.7<br>
                    ><br>
                    > does the above work for you?  oe<br>
                    ><br>
                    > On Thu, Oct 4, 2018 at 12:56 AM Asad Sayeed
                    <<a href="mailto:asayeed@mbl.ca" target="_blank"
                      moz-do-not-send="true">asayeed@mbl.ca</a>>
                    wrote:<br>
                    >> Hi,<br>
                    >><br>
                    >> I've started trying to use the cluster. 
                    It's great but my current task<br>
                    >> seems to be impossible there, so I am
                    running it for longer on local<br>
                    >> machines.  The most recent issue is that I
                    wanted to run a neural<br>
                    >> semantic role labeller that requires a
                    recent tensorflow that runs on<br>
                    >> python2.7.  I was only able to find a
                    tensorflow that runs on python3,<br>
                    >> and my attempts to use a singularity
                    container failed on abel even when<br>
                    >> I tried to package the container on my
                    local system with the same<br>
                    >> singularity version that is on abel.  Also
                    virtualenv and user-level<br>
                    >> package install did not work for me because
                    of an incompatible glibc<br>
                    >> library. Would there be any chance of
                    getting a tensorflow for python2.7<br>
                    >> working on abel, along with tensorflow_hub?<br>
                    >><br>
                    >> The purpose is to perform SRL on about 10
                    million sentences, which<br>
                    >> should take a day if I can run 400
                    processes with 10GB memory each (to<br>
                    >> hold glove vectors).  I may in the future
                    have even larger SRL-related<br>
                    >> tasks.<br>
                    >><br>
                    >> Yours,<br>
                    >> --Asad.<br>
                    >><br>
                    <br>
                  </blockquote>
                </div>
              </div>
            </blockquote>
            <br>
          </div>
        </blockquote>
      </div>
    </blockquote>
    <br>
  </body>
</html>