Installation#

Hardware Installation#

The module is not recognized by the host system.

There are a few possible explanations why the MX3 M.2 may not be detected by a system.

Kernel Module Needs Inserting

Sometimes, installing the driver ( memx-drivers deb package, or manual installer script ) will build the kernel module but not automatically insert it. Either reboot your system, at which time the module will autoload on next boot, or manually insert the module with the following command:

sudo modprobe -v memx_cascade_plus_pcie

ARM System Needs Device Tree Modification

Some ARM boards require device tree overlays or other modifications to support the MX3 M.2 card. The sudo mx_arm_setup command in the install docs will do these tweaks for you on currently supported platforms. Support for more boards out-of-the-box will continue to grow over time. Please contact MemryX for support if your ARM system is having install issues.

The included MemryX heat sink does not fit in my host system.

If the MemryX heat sink does not fit properly into the host system, it is necessary to find an alternative heat sink for the M.2 module. Some host systems will come with pre-existing flat heat sinks, which will often be sufficient for the module. We recommend transferring the thermal pads included with the MemryX heat sink to the alternate heat sink of your choice for optimal performance and thermal management.

Software Installation#

Installing a supported Python with uv

The Python packages in the MemryX SDK (NeuralCompiler, etc.) require Python 3.9 through 3.12. If you have an older or newer version of Python, you’ll need to install a supported version first.

We recommend using uv, a Python version manager that makes it easy to install and switch between multiple Python versions.

For example, to install Python 3.12 and install the `memryx` package into a venv:

1. Install uv*

# install uv
curl -LsSf https://astral.sh/uv/install.sh | sh

# now either restart your shell or call this
source $HOME/.local/bin/env

2. Install Python 3.12 and make a venv

# install 3.12
uv python install 3.12

# make a venv at ~/mx
uv venv -p 3.12 ~/mx

# activate the venv
source ~/mx/bin/activate

2.5. Make sure you have cmake installed

sudo apt install cmake

3. Install the memryx package

uv pip install --extra-index-url https://developer.memryx.com/pip memryx

4. Return to the installation guide

You have successfully installed the SDK Python Tools. You can now return to the installation guide. No need to install the memryx pip package again.

Python: No module named venv

You need to install venv for your OS. For Ubuntu, you can use sudo apt install python3.10-venv (or python3.12-venv, etc.). Then, retry the command to create the virtual environment.

Note

As a reminder, MemryX explicitly supports Python 3.9, 3.10, 3.11, and 3.12. Python 3.13 support is coming soon.