<div dir="ltr"><div dir="ltr"><div>Regarding tensorflow you don't need to compile from source.<br><pre class="gmail-lang-bsh"><code class="gmail-devsite-terminal"><span class="gmail-pln">pip install tf</span><span class="gmail-pun">-</span><span class="gmail-pln">nightly</span><span class="gmail-pun">-</span><span class="gmail-pln">gpu</span></code></pre></div>Should get it done. I think that's what I've done, but it's been few weeks ago, so try it out and if it doesn't work I'll try to debug it.</div><div>Notice that you'll have to uninstall it and install the regular version when you switch back to the other GPUs. <br></div><div><br></div><div>Not sure regarding pytorch, I haven't tried to install it yet.</div><div><br></div><div>Yotam.</div><div><br></div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Sun, Mar 10, 2019 at 2:24 PM Yichong Xu <<a href="mailto:yichongx@cs.cmu.edu">yichongx@cs.cmu.edu</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">



<div style="overflow-wrap: break-word;">
It seems like tensorflow does not support cuda10 right now - it has to be installed from source.
<div>But I’m mainly using pytorch though and the version with cuda10 does not run either.</div>
<div>Plus, I tried the original cuda example and it cannot find the gpu either:</div>
<div>
<div>(base) yichongx@gpu10$ ls</div>
<div>Makefile  readme.txt  simplePrintf  simplePrintf.cu  simplePrintf.o</div>
<div>(base) yichongx@gpu10$ ./simplePrintf</div>
<div>CUDA error at ../../common/inc/helper_cuda.h:744 code=999(cudaErrorUnknown) "cudaGetDeviceCount(&device_count)"</div>
<div>(base) yichongx@gpu10$</div>
</div>
<div><br>
</div>
<div><br>
</div>
<div><br>
<div>
<div style="color:rgb(0,0,0);font-family:Helvetica;font-size:12px;font-style:normal;font-variant-caps:normal;font-weight:normal;letter-spacing:normal;text-align:start;text-indent:0px;text-transform:none;white-space:normal;word-spacing:0px">
<i>Thanks,</i></div>
<div style="color:rgb(0,0,0);font-family:Helvetica;font-size:12px;font-style:normal;font-variant-caps:normal;font-weight:normal;letter-spacing:normal;text-align:start;text-indent:0px;text-transform:none;white-space:normal;word-spacing:0px">
<i>Yichong</i></div>
<div style="color:rgb(0,0,0);font-family:Helvetica;font-size:12px;font-style:normal;font-variant-caps:normal;font-weight:normal;letter-spacing:normal;text-align:start;text-indent:0px;text-transform:none;white-space:normal;word-spacing:0px">
<br>
</div>
<br class="gmail-m_1836210367494301620Apple-interchange-newline">
</div>
<div><br>
<blockquote type="cite">
<div>On Mar 10, 2019, at 9:52 AM, Yotam Hechtlinger <<a href="mailto:yhechtli@andrew.cmu.edu" target="_blank">yhechtli@andrew.cmu.edu</a>> wrote:</div>
<br class="gmail-m_1836210367494301620Apple-interchange-newline">
<div>It's not the same cuda version on GPU 10 and the rest, I think different version of tensorflow has to be installed.
<div><br>
</div>
<div>Check your tensorflow version and if it supports the cuda version on gpu10.</div>
<div><br>
</div>
<div><br>
<br>
On Saturday, March 9, 2019, Predrag Punosevac <<a href="mailto:predragp@andrew.cmu.edu" target="_blank">predragp@andrew.cmu.edu</a>> wrote:<br>
<blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
<div dir="auto">Try CUDA 10.0 instead of 10.1</div>
<div class="gmail_extra"><br>
<div class="gmail_quote">On Mar 9, 2019 5:28 PM, Yichong Xu <<a href="mailto:yichongx@cs.cmu.edu" target="_blank">yichongx@cs.cmu.edu</a>> wrote:<br type="attribution">
<blockquote style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
<div dir="auto">Same issue here.<br>
<br>
<div dir="ltr">From my iPhone</div>
<div dir="ltr"><br>
On Mar 9, 2019, at 4:01 PM, Emre Yolcu <<a href="mailto:eyolcu@andrew.cmu.edu" target="_blank">eyolcu@andrew.cmu.edu</a>> wrote:<br>
<br>
</div>
<blockquote>
<div dir="ltr">
<div>
<p>Hi,</p>
<div> <br class="gmail-m_1836210367494301620webkit-block-placeholder">
</div>
<p>Right now on gpu10 `nvcc --version` and `nvidia-smi` seem to work, but `python -c ‘import torch; print(torch.cuda.is_available())’` prints False. Is anybody running into the same issue?</p>
<div> <br class="gmail-m_1836210367494301620webkit-block-placeholder">
</div>
<p>Emre</p>
</div>
</div>
</blockquote>
</div>
</blockquote>
</div>
<br>
</div>
</blockquote>
</div>
</div>
</blockquote>
</div>
<br>
</div>
</div>

</blockquote></div>