Smartdispatch
An easy to use job launcher for supercomputers with PBS compatible job manager.
Install / Use
/learn @SMART-Lab/SmartdispatchREADME
Smart Dispatch
An easy to use job launcher for supercomputers with PBS compatible job manager.
Features
- Launch multiple jobs with a single line.
- Automatically generate combinations of arguments. (see examples)
- Automatic resources management. Determine for you the optimal fit for your commands on nodes.
- Resume batch of commands.
- Easily manage logs.
- Advanced mode for complete control.
Installing
pip install git+https://github.com/SMART-Lab/smartdispatch
Usage
See smart-dispatch --help
Output and error logs in are saved in : ./SMART_DISPATCH_LOGS/{batch_id}/logs/.
Examples
Launch job
smart-dispatch -q qtest@mp2 launch python my_script.py 2 80 tanh 0.1
Will launch python my_script.py 2 80 tanh 0.1 on the queue qtest@mp2.
Launch batch of jobs
Automatically generate commands from combinations of arguments.
smart-dispatch -q qtest@mp2 launch python my_script.py [1 2] 80 [tanh sigmoid] 0.1
Will generate 4 different commands and launch them on the queue qtest@mp2:
python my_script.py 1 80 sigmoid 0.1
python my_script.py 1 80 tanh 0.1
python my_script.py 2 80 sigmoid 0.1
python my_script.py 2 80 tanh 0.1
Another possiblility is to generate argument from a range.
smart-dispatch -q qtest@mp2 launch python my_script.py [1:4]
Will generate:
python my_script.py 1
python my_script.py 2
python my_script.py 3
You can also add a step size to the range as the 3rd argument.
smart-dispatch -q qtest@mp2 launch python my_script.py [1:10:2]
Will generate:
python my_script.py 1
python my_script.py 3
python my_script.py 5
python my_script.py 7
python my_script.py 9
Resuming job/batch
smart-dispatch -q qtest@mp2 resume {batch_id}
Jobs that did not terminate properly, for example, it exceeded the walltime, can be resumed using the {batch_id} given to you upon launch. Of course, all this assuming your script is resumable.
Note: Jobs are always in a batch, even if it's a batch of one.
Related Skills
node-connect
342.5kDiagnose OpenClaw node connection and pairing failures for Android, iOS, and macOS companion apps
claude-opus-4-5-migration
85.3kMigrate prompts and code from Claude Sonnet 4.0, Sonnet 4.5, or Opus 4.1 to Opus 4.5
frontend-design
85.3kCreate distinctive, production-grade frontend interfaces with high design quality. Use this skill when the user asks to build web components, pages, or applications. Generates creative, polished code that avoids generic AI aesthetics.
model-usage
342.5kUse CodexBar CLI local cost usage to summarize per-model usage for Codex or Claude, including the current (most recent) model or a full model breakdown. Trigger when asked for model-level usage/cost data from codexbar, or when you need a scriptable per-model summary from codexbar cost JSON.



