A simple program scheduler for your code on different devices.

Let the machine move!

Putting the machine into sleep is a disrespect for time.

• PyYAML==6.0
• nvidia-ml-py (pynvml only for runit_based_on_detected_memory.py)

We provides 3 scripts for different ways to run jobs.

• runit_with_exclusive_gpu.py: One GPU can only be used by one job at a time.
• runit_based_on_memory：One GPU can be used by many job at a time based on the memory usage.
• runit_based_on_detected_memory.py: Use pynvml for detecting the total memory usage of each GPU. But this may not be suitable for scenarios where the memory used by a running GPU application is unstable.

$ python run_it.py --config ./examples/config.yaml
$ python run_it.py --max-workers 3 --config ./examples/config.yaml

graph TD
    A[Start] --> B[Read Configuration and Command Pool]
    B --> C[Initialize Shared Resources]
    C --> |Maximum number of requirements met| D[Loop Until All Jobs Done]
    D --> E[Check Available GPUs]
    E -->|Enough GPUs| F[Run Job in Separate Process]
    E -->|Not Enough GPUs| G[Wait and Retry]
    F --> H[Job Completes]
    F --> I[Job Fails]
    H --> J[Update Job Status and Return GPUs]
    I --> J
    G --> D
    J -->|All Jobs Done| K[End]
    C -->|Maximum number of requirements not met| L[Terminate Workers]
    L --> M[Shutdown Manager and Join Pool]
    M --> K

• @BitCalSaul: Thanks for the positive feedbacks!
  • #3
  • #2
  • #1
• https://www.jb51.net/article/142787.htm
• https://docs.python.org/zh-cn/3/library/subprocess.html
• https://stackoverflow.com/a/23616229
• https://stackoverflow.com/a/14533902