Setup Cosine Similarity for Stand-Alone Runs

1. Create a virtual environment and install dependencies

Using Conda

$ conda env list
$ conda create -n fpga python=3.6
$ conda activate fpga
(fpga)$ pip install -r /opt/xilinx/apps/graphanalytics/requirements.txt

Using local python

  • At least Python version 3.6 is required, install or upgrade python using your package manager

  • Create and activate a new virtual environment and install all required packages

$ python3 -m venv fpga
$ source fpga/bin/activate
(fpga)$ pip install -r /opt/xilinx/apps/graphanalytics/requirements.txt

2. Copy examples to user accessible directory

(fpga)$ cp -r /opt/xilinx/apps/graphanalytics/cosinesim/1.5.1/examples cosinesim-examples
(fpga)$ cd cosinesim-examples/python

Setup for Jupyter Notebooks

Note

If running Jupyter Notebook(s), read this section and follow instructions in the corresponding Notebook Demo pages to run.

The stand-alone cosine similarity jupyter notebooks run on the same server (local) that has the Alveo cards installed, but can be launched locally or remotely from a system that is on the same network. Consequently, there are two ways to run the notebooks:

Method 1: Run on the local server

  • Simply start the jupyter server as:

(fpga)$ ./run.sh jupyter notebook

  • Navigate to the Notebook under current directory that you want to run

Method 2: Run from a remote machine

  • JupyterHub is not part of the requirements.txt dependency list. Install JupyterHub on the local machine as shown here

  • Start a JupyterHub server on the local machine as:

(fpga)$ ./run.sh jupyterhub

  • Open a web browser on your remote machine, enter the IP address and port of the Jupyterhub server

  • Navigate to the Notebook under current directory that you want to run