A simple dash-based GPU monitor that uses web pages to continuously view GPU usage.
- Use the webpage to continuously view basic GPU usage, including GPU usage, remaining graphics memory, temperature, and fan speed
- Check who is using the GPU, including user, process ID, process creation time, process creation commands, etc. For details, see the description below.
Bifrost relies mainly on the following:
- Python 3.6+
- dash
- psutil
- pandas
- numpy
- nvidia-ml-py3
You can install these dependencies with the following command:
pip install dash psutil pandas numpy nvidia-ml-py3
You can also use requirements.txt
to install them:
pip install -r requirements.txt
python app.py
or using gunicorn:
gunicorn -c config.py app:server
Then open http://locahost:8150 or http://{your_ip}:8150 in your browser. Bazinga!
gpu
: GPU IDpid
: The process IDused_gpu_mem
: The GPU memory used by the processusername
: The owner of the processname
: The Process namecreate_time
: The process creation timestatus
: The process statuscpu_percent
: CPU utilization of the processcpu_num
: Which CPU this process is currently running onmemory_percent
: Memory utilization of the processnum_threads
: The number of threads currently used by this processcmdline
: The command line this process has been called
MIT.