parvvaresh / Monitor-GPU-usage

This repository contains a Python script that monitors GPU usage and logs the data into a CSV file at regular intervals. The script leverages nvidia-smi to query GPU statistics and runs in the background using the screen utility.

GPU Usage Monitor

This repository contains a Python script that monitors GPU usage and logs the data into a CSV file at regular intervals. The script leverages nvidia-smi to query GPU statistics and runs in the background using the screen utility.

Project Structure

.
├── _gpu_usage
│   ├── usgae_gpu.sh
├── report_server
    ├── report_server.py
├── gpu_usage.py
└── README.md

Files and Their Functions

• _gpu_usage/install_screen.py: Installs the screen utility.
• _gpu_usage/run_commands.py: Executes necessary commands to set up and run the monitoring script.
• _gpu_usage/run_in_screen.py: Runs the monitoring script in a detached screen session.
• _gpu_usage/write_sh_file.py: Creates a shell script that logs GPU usage data.
• gpu_usage.py: Main script that coordinates the process.

will be update

Prerequisites

• Python 3.x
• NVIDIA GPU with nvidia-smi installed
• screen utility (the script will install it if not present)

Installation

1. Clone the Repository:

   git clone https://github.com/parvvaresh/Monitor-GPU-usage
   cd Monitor-GPU-usage

2. Run the Script:

   python3 gpu_usage.py

   The script will prompt you to enter the interval (in seconds) at which you want to log the GPU usage data.

Usage

1. Enter the Logging Interval: When prompted, specify how frequently you want to save the GPU usage data (in seconds).

2. Background Logging: The script will:

   • Make the logging shell script executable.
   • Install screen if it is not already installed.
   • Run the logging shell script in a detached screen session named usage_gpu.

3. Access the Log File: The GPU usage data will be logged in a file named result.csv in the same directory as the script.

report server

1. Clone the Repository:

   git clone https://github.com/parvvaresh/Monitor-GPU-usage
   cd Monitor-GPU-usage

2. Use class:

   from report_server.report_server import report_server
   rs = report_server(df) # data frame who save it in your server
   rs.get_report_plot()
   rs.get_text_report()

Contributing

Contributions are welcome. Please fork the repository and submit a pull request for review.

License

This project is licensed under the MIT License. See the LICENSE file for details.