All - Activity - jeremy - DeepTalk - Deep Learning Community

Lambda Stack seems to be wiped out after Ubuntu update

Ivo! Hello again!
Interesting to see that nvidia-smi isn’t being installed.
Looks like it’s a part of the nvidia-utils-460 package - let’s try installing that manually:

sudo apt install nvidia-utils-460

Maybe that will tell us why it didn’t install with lambda-stack-cuda.


Ethernet interface IDs change and break network

This name change is because of the way SystemD enumerates devices - I’m not sure why this would be the intended behavior.
Put the following in /etc/udevd/rules.d/10-ethernet.rules:

SUBSYSTEM=="net", ACTION=="add", ATTR{address}=="01:23:45:67:89:ab", NAME="eth1"

The address above will be the MAC o…


Updating cudatoolkit to 10.1

This is probably one of the Python 2 dev packages.
Try removing them -

sudo apt-get remove python-*dev

Then try the release upgrade again.


Cannot apt --fix, potentially conflicting packages

Try this:

sudo apt install --fix-broken

Then reboot.
We have seen this issue crop up with other users recently, and are current looking into the potential causes.


Reset Root Password

Steps to Create a Bootable USB of Ubuntu

  1. Create a bootable USB of Ubuntu (or any flavor of Linux)
    We discuss creating a bootable USB here.
  2. Boot into the live image
    Power the machine on and press Del when you see the Asus logo.
    Then press F8 to open the boot menu. Select your USB drive.
    From here, select Try Ubuntu without Installing…

Path for labmda stack

The following will list the paths of all of the files associated with the Caffe installation:

dpkg -L python3-caffe-cuda

How to turn off the backlight on keyboard

You can run the following to be able to control the keyboard LEDs via the (Fn+) +, -, * and / keys on the number pad of the keyboard:

git clone https://github.com/tuxedocomputers/tuxedo-keyboard  
cd tuxedo-keyboard  
make && sudo make dkmsinstall && \  
sudo modprobe tuxedo_keyboard

Fn+* will then allow...


Problem with CuDNN initialization

Try adding the following after importing Tensorflow:

for device in tf.config.experimental.list_physical_devices('GPU'):
    tf.config.experimental.set_memory_growth(device, True)

Dist-upgrade does not upgrade to Cuda 11.1

The latest version of the CUDA toolkit we support on Ubuntu 18.04 is version 10.0.130.
If you want to use version 11.1, you’ll need to upgrade to Ubuntu 20.04.
You can do that like so:

sudo sed -i '/^AutomaticLoginEnable/ s/true/false/' /etc/gdm3/custom.conf &&\  
sudo apt -y update && sudo apt -y ...

Tensorflow in virtual environment on GPU Workstation

Assuming you’re on Ubuntu 20.04, simply run the following:

sudo apt -y install libcudart10.1

Hyperplane Setup

Mellanox OFED, NVIDIA Peer Memory, NVIDIA Driver, CUDA, cuDNN & Docker

Download the Mellanox OFED tarball
and copy it to /tmp. We’ll need it later:

cp MLNX_OFED_LINUX-5.1-2.3.7.1-ubuntu18.04-x86_64.tgz /tmp

Trying to find TF version and getting "Invalid MIT-MAGIC-COOKIE-1..."

It seems that the Invalid MIT-MAGIC-COOKIE-1 key message doesn’t pop up at all when using lightdm.
To switch to lightdm, run:

sudo apt -y install lightdm

Once that installs, run the following, and select lightdm:

sudo dpkg-reconfigure lightdm

Then reboot.


Can't build keras model after Tensorflow update from 2.2 to 2.3

See the new workaround. Also, try

pip3 uninstall tensorflow

to revert back to the version installed by Lambda Stack.
If the error message doesn’t change, try adding the following after importing tensorflow:

for device in tf.config.experimental.list_physical_devices('GPU'):
    tf.config.experiment...

Cuda Error with Tensorflow > 2.0

I’ve only run into this issue when I two models training on the same GPU.
Was the output any different when you gave the memory growth approaches a try?

devices = tf.config.experimental.list_physical_devices('GPU')
for device in devices:
    tf.config.experimental.set_memory_growth(devices, True)

How to change keyboard backlight color

Though I did get the keyboard backlights to work by ACPI kernel parameters as Harenbrs suggested, I was no longer able to “resume from suspend” without a hard reboot.
I compiled and installed the “tuxedo-keyboard” kernel module:

git clone https://github.com/tuxedocomputers/tuxedo-keyboard
cd tuxed...

External Monitors No longer working

Try running the following to revert to the previous version of the NVidia driver, then reboot:

dpkg -l | awk '$2 ~ /-440/ { printf "%s=440.100-0ubuntu0.18.04.1\n", $2 }' |sudo xargs apt -y install --allow-downgrades

If your display doesn’t come back, could you run sudo nvidia-bug-report.sh and...


Error importing tensorflow after upgrade to v2

Run the following to remove all packages installed via pip, and revert back to the Python packages installed by Lambda Stack:

sudo rm -rf {~/.local,/usr/local}/{bin/pip,lib/python}*

You should then be able to import tensorflow successfully.