Executions
An execution is one run of one version. It records everything needed to know what ran: the version, the environment, the executable, the number of MPI ranks, extra arguments and environment variables, and whether profiling was on.
States
Section titled “States”| State | Meaning |
|---|---|
| Pending | Created, not yet handed to the machine. |
| Submitted | The machine accepted it and is starting the engine. |
| Running | The engine is running; thermo and resource use are being read. |
| Completed | The engine exited normally. |
| Failed | It exited with an error or could not start. The page shows why. |
| Cancelled | You cancelled it and the process has exited. |
A run can be cancelled while submitted or running.
The run page
Section titled “The run page”| Tab | Contents |
|---|---|
| Overview | Timesteps per second, wall time, CPU, peak memory, time left, profiles. |
| Thermo | Every thermo column, plotted as it is written. One chart per run command. |
| Speed | Seconds per step over time, the slowest interval, time left for the current run. |
| Resources | CPU, memory and threads across the process tree, and each MPI rank’s CPU over time. |
| Profiling | Flame graph and hotspots when the run was profiled. |
| Logs | The engine’s log, standard output and standard error, followed live. |
| Artifacts | Every file the run wrote: logs, dumps, restart files. |
| Configuration | The exact input, executable, ranks, arguments and environment. |
The copy of the input that runs is made to flush its thermo output on every line (for LAMMPS,
thermo_modify flush yes is added), so the charts update as soon as a line is computed. The saved
version is not changed.
Where runs happen
Section titled “Where runs happen”Each run has its own working directory: inside verlet-agent’s data folder on your computer, under
~/.verlet/work on a remote machine. The input is copied there before the engine starts.
Runs are detached. A remote run keeps going if your laptop sleeps, the connection drops or verlet is closed; on reconnect, its progress and final state catch up.
Outputs
Section titled “Outputs”Local outputs are already on your computer. For a remote run, Download all outputs on the Artifacts tab copies every file once it has finished; single files download at any time.
Comparing runs
Section titled “Comparing runs”Select two runs on Executions and press Compare, or open the comparison from a run page to compare it with the one before. It shows speed and timing differences, thermo curves overlaid, the timing breakdown and, for profiled runs, how the hotspots moved. It reads as “B relative to A”; green means B improved.