RunPod GPU Tester

The RunPod-GPU-Tester is a diagnostic tool designed to help users and tech support diagnose and troubleshoot GPU-related issues in RunPod environments. This Python script collects essential system and environment information, performs basic CUDA operations to test each GPU's functionality, and generates a diagnostic report that users can share with RunPod tech support for further assistance.

Features

• Collects system and GPU environment information.
• Performs CUDA operations to test the functionality of all available GPUs.
• Generates a detailed diagnostic report (gpu_diagnostics.json) including any errors encountered during the tests.

Prerequisites

• Python 3.6 or later.
• Access to a RunPod environment with GPUs (obviusly) .

Quick Start

To use the RunPod-GPU-Tester, follow these simple steps:

1. Download the script directly into your RunPod environment using the following command:

   wget https://github.com/kodxana/RunPod-GPU-Tester/raw/main/tester.py -O tester.py
   

2. Run the script using Python:

   python tester.py
   

3. After the script completes, it will generate a gpu_diagnostics.json file in the /workspace directory. Share this file with RunPod tech support for further analysis and assistance.

Output

The script will save the diagnostic information in a file named gpu_diagnostics.json in the /workspace directory. This file contains:

• System information including CUDA version, PyTorch version, and GPU details.
• Results from the CUDA functionality tests for each GPU.
• The machine ID associated with the current RunPod pod .

Contributing

Contributions to the RunPod-GPU-Tester are welcome! Please feel free to submit pull requests or open issues to suggest improvements or report bugs.

Support

If you encounter any issues or have questions about using the script, please open an issue in this GitHub repository.