Fuzzball Documentation
Toggle Dark/Light/Auto mode Toggle Dark/Light/Auto mode Toggle Dark/Light/Auto mode Back to homepage
  2. /
  3. Workflow Examples
  4. /
  5. Open Source Examples
  6. /
  7. LAMMPS
  8. /
  9. Executing a LAMMPS Workflow

Executing a LAMMPS Workflow

The Fuzzfile below runs a standard Lenard-Jones 3D melt experiment using LAMMPS. The workflow generates output file log.lammps which contains simulation and benchmark results which can be saved through data egress to a path defined in an S3 URI. This workflow uses a LAMMPS container from Nvidia’s NGC catalog and runs on a single node with a GPU.

version: v1
volumes:
  lammps-data-volume:
    name: lammps-data-volume
    reference: volume://user/ephemeral
    ingress:
      - source:
          uri: https://www.lammps.org/inputs/in.lj.txt
        destination:
          uri: file://in.lj.txt
    # egress:
    #   - source:
    #       uri: file://log.lammps
    #     destination:
    #       uri:
    #       secret:
jobs:
  run-lammps:
    image:
      uri: docker://nvcr.io/hpc/lammps:29Sep2021
    env:
      - LAMMPS_INPUT_FILE_PATH=/data/in.lj.txt
    command:
      - /bin/sh
      - '-c'
      - >-
        lmp -k on g 1 -sf kk -pk kokkos cuda/aware on neigh full comm device binsize 2.8 -var x 8 -var y 4 -var z 8
        -in ${LAMMPS_INPUT_FILE_PATH}        
    cwd: /data
    resource:
      cpu:
        cores: 2
        affinity: NUMA
      memory:
        size: 14GiB
      devices:
        nvidia.com/gpu: 1
    mounts:
      lammps-data-volume:
        location: /data

You can run this workflow either through the GUI or the CLI.

Please select either the GUI or CLI tab to see the appropriate instructions for your environment.

If you click “Workflow Editor” and “Create New”, you will see a blank page in the workflow editor.

Fuzzball create new workflow section

Now you can either click the ellipses (...) menu in the lower right and select “Edit YAML” or simply press e on your keyboard. An editor with a Fuzzfile stub will appear.

Fuzzball editor newly opened

You can delete the current contents and copy and paste the workflow definition of from above.

Fuzzball editor copied and pasted LAMMPS workflow

Now pressing “save” will return you to the interactive workflow editor. You will now see a job named run-lammps instead of a blank editor page. The Fuzzball GUI will automatically validate the yaml file for syntax errors.

Fuzzball workflow editor with LAMMPS

Submitting your workflow to Fuzzball with the GUI is easy. Simply press the triangular “Start Workflow” button in the lower right corner of the workflow editor. You will be prompted to provide an optional descriptive name for your workflow.

Fuzzball name workflow and submit screen

Now you can click on “Start Workflow” in the lower right corner of the dialog box and your workflow will be submitted. If you click “Go to Status” you can view the workflow status page. The screenshot below shows the status page for a hello world workflow submission.

Fuzzball workflow status page

To retrieve logs produced by this workflow, select a job within the workflow such as make-blast-database, and click the “Logs” option on the right.

Fuzzball workflow dashboard showing logs from the run-lammps job

To run this workflow through the CLI you will need access to the Fuzzball CLI. You can install it using the Fuzzball CLI installation instructions.

First, you can create a Fuzzfile lammps-gpu-ngc.fz with the contents above using the text editor of your choice.

You can start start this workflow using the CLI by running the following command:

$ fuzzball workflow start lammps-gpu-ngc.fz
Workflow "45d81a83-dd54-4052-afab-09dd3216abbe" started.

You can monitor the workflow’s status by running the following command:

$ fuzzball workflow describe <workflow uuid>
Name:      lammps-gpu-ngc.fz
Email:     bphan@ciq.co
UserId:    e554e134-bd2d-455b-896e-bc24d8d9f81e
Status:    STAGE_STATUS_FINISHED
Created:   2024-06-21 09:23:58AM
Started:   2024-06-21 09:23:58AM
Finished:  2024-06-21 09:26:44AM
Error:


Stages:
KIND     | STATUS   | NAME                                    | STARTED               | FINISHED
Workflow | Finished | 45d81a83-dd54-4052-afab-09dd3216abbe    | 2024-06-21 09:23:58AM | 2024-06-21 09:26:44AM
Volume   | Finished | lammps-data-volume                      | 2024-06-21 09:23:58AM | 2024-06-21 09:24:16AM
Image    | Finished | docker://nvcr.io/hpc/lammps:29Sep2021   | 2024-06-21 09:23:58AM | 2024-06-21 09:24:15AM
File     | Finished | https://www.lammps.org/inputs/in.lj.txt | 2024-06-21 09:24:31AM | 2024-06-21 09:24:35AM
         |          | ->...                                   |                       |
Job      | Finished | run-lammps                              | 2024-06-21 09:25:46AM | 2024-06-21 09:26:09AM
File     | Finished | file://log.lammps ->                    | 2024-06-21 09:26:24AM | 2024-06-21 09:26:28AM
         |          | s3://co-ciq-misc-supp...                |                       |

You can view outputs logged by the workflow using the fuzzball workflow log command and provide the workflow UUID and job name. For example, executing the following command, will output logs from job run-lammps in the workflow:

$ fuzzball workflow logs <workflow uuid> run-lammps
LAMMPS (29 Sep 2021)
KOKKOS mode is enabled (src/KOKKOS/kokkos.cpp:97)
  will use up to 1 GPU(s) per node
  using 1 OpenMP thread(s) per MPI task
Lattice spacing in x,y,z = 1.6795962 1.6795962 1.6795962
Created orthogonal box = (0.0000000 0.0000000 0.0000000) to (268.73539 134.36770 268.73539)
  1 by 1 by 1 MPI processor grid
Created 8192000 atoms
  using lattice units in orthogonal box = (0.0000000 0.0000000 0.0000000) to (268.73539 134.36770 268.73539)
  create_atoms CPU = 3.047 seconds
Neighbor list info ...
  update every 20 steps, delay 0 steps, check no
  max neighbors/atom: 2000, page size: 100000
  master list distance cutoff = 2.8
  ghost atom cutoff = 2.8
  binsize = 2.8, bins = 96 48 96
  1 neighbor lists, perpetual/occasional/extra = 1 0 0
  (1) pair lj/cut/kk, perpetual
      attributes: full, newton off, kokkos_device
      pair build: full/bin/kk/device
      stencil: full/bin/3d
      bin: kk/device
Setting up Verlet run ...
  Unit style    : lj
  Current step  : 0
  Time step     : 0.005
Per MPI rank memory allocation (min/avg/max) = 1181.0 | 1181.0 | 1181.0 Mbytes
Step Temp E_pair E_mol TotEng Press
       0         1.44   -6.7733681            0   -4.6133683   -5.0196694
     100   0.75927734    -5.761232            0   -4.6223161   0.19102612
Loop time of 1.92832 on 1 procs for 100 steps with 8192000 atoms
Performance: 22402.943 tau/day, 51.859 timesteps/s
72.4% CPU use with 1 MPI tasks x 1 OpenMP threads
MPI task timing breakdown:
Section |  min time  |  avg time  |  max time  |%varavg| %total
---------------------------------------------------------------
Pair    | 0.16802    | 0.16802    | 0.16802    |   0.0 |  8.71
Neigh   | 0.29642    | 0.29642    | 0.29642    |   0.0 | 15.37
Comm    | 0.072074   | 0.072074   | 0.072074   |   0.0 |  3.74
Output  | 0.00063816 | 0.00063816 | 0.00063816 |   0.0 |  0.03
Modify  | 1.3215     | 1.3215     | 1.3215     |   0.0 | 68.53
Other   |            | 0.06967    |            |       |  3.61
Nlocal:    8.19200e+06 ave   8.192e+06 max   8.192e+06 min
Histogram: 1 0 0 0 0 0 0 0 0 0
Nghost:        727553.0 ave      727553 max      727553 min
Histogram: 1 0 0 0 0 0 0 0 0 0
Neighs:         0.00000 ave           0 max           0 min
Histogram: 1 0 0 0 0 0 0 0 0 0
FullNghs:  6.15293e+08 ave 6.15293e+08 max 6.15293e+08 min
Histogram: 1 0 0 0 0 0 0 0 0 0
Total # of neighbors = 6.1529347e+08
Ave neighs/atom = 75.109066
Neighbor list builds = 5
Dangerous builds not checked
Total wall time: 0:00:13
gdoc_arrow_left_alt LAMMPS Saving LAMMPS Results gdoc_arrow_right_alt