Help building Tensorflow 2.1 for Xavier

Hi I am following the instructions to build TensorFlow 2.1 for Python 3.7 (I will share the wheel after) but after 24 hours of building it crashes and says out of space. Yes my device is really out of space its like 39mb left (i dont know who thought it would be a good idea to put only 32gb of space as default)

Problem is i cannot delete anything else. The OpenCV 4.1.0 for python3 itself is 10gb and even after temporarily moving the OpenCV folder to a USB, i still dont have enough space. Right now it is consuming all of the 9gb free disk space that i have left.

I ordered an NVME SSD, could anyone pelase confirm that if i git pull the tensorflow code onto the NVME drive and i used bazel to compile it, would it then be consuming the space required on the NVME drive instead of the boot drive?

I’m not an expert in linux but i want to know before i go through the steps of opening it and installing the NVME

Thanks!

Hi,

YES.
If you build TensorFlow on an extra SSD, the output file will be stored to the sub-folder which should also located at the SSD.

Looking forward to your sharing.
Thanks.

Thanks! Will confirm, it is still compiling on the boot drive if it fails tomorrow then i will install it on the SSD.

I also have another question unrelated to this issue but about the build. I will start a new thread

Hi - where did you find the instructions? Are there specific instructions for the Xavier vs. just building on an x86?

That the part i dont know yet. I’m still trying to cpompile TF2.1 but its failing because of errors.

In parallel i setup a Ubuntu desktop and im compiling TF1.14 now so i’ll see how that goes. In my other post i posted a link for instructions someone wrote down for TF1.13 for Xavier. Apparently it only needs a couple of hacks in the build files. My compilation of TF1.14 will confirm this