mirror of
https://github.com/hanzo-ml/trainer.git
synced 2026-09-24 00:31:34 +00:00
Amir IbrahimandGitHub
49f4eaadd2
* test: add Flux framework integration and e2e tests Signed-off-by: Amir380-A <[email protected]> * fix: fix integration and e2e Signed-off-by: Amir380-A <[email protected]> * fix: EOF linting issue Signed-off-by: Amir380-A <[email protected]> * chore: add logs for E2E ClusterTrainingRuntimes in Github Actions Signed-off-by: Amir380-A <[email protected]> * feat: add Flux runtime to Helm chart Signed-off-by: Amir380-A <[email protected]> * fix: fix EOF linting issue Signed-off-by: Amir380-A <[email protected]> * fix: install Flux runtime only for E2E tests Signed-off-by: Amir380-A <[email protected]> * chore: regenerate Helm chart docs and generated assets Signed-off-by: Amir380-A <[email protected]> * fix: retry failed deepspeed CI dependency Signed-off-by: Amir380-A <[email protected]> * revert back to the previous commit Signed-off-by: Amir380-A <[email protected]> * fix(e2e): install Flux runtime only during E2E setup Signed-off-by: Amir380-A <[email protected]> * fix: fix linting issue Signed-off-by: Amir380-A <[email protected]> * fix: fix runtimeRef for CI Signed-off-by: Amir380-A <[email protected]> * fix: fix flux examples runtimeref Signed-off-by: Amir380-A <[email protected]> --------- Signed-off-by: Amir380-A <[email protected]> Signed-off-by: Amir Ibrahim <[email protected]>
27 lines
952 B
YAML
27 lines
952 B
YAML
# This example deploys the LAMMPS Molecular Dynamic Simulator
|
|
# with MPI orchestrated by the Flux workload manager on 4 nodes.
|
|
# This is "interactive" mode, so you can shell in to execute lmp.
|
|
# See lammps-train-job.yaml for more details on the setup, and
|
|
# instructions below for running lammps.
|
|
apiVersion: trainer.kubeflow.org/v1alpha1
|
|
kind: TrainJob
|
|
metadata:
|
|
name: lammps-flux-interactive
|
|
spec:
|
|
# Reference the pre-defined runtime by name
|
|
runtimeRef:
|
|
name: flux-distributed
|
|
trainer:
|
|
numNodes: 4
|
|
image: ghcr.io/converged-computing/metric-lammps:latest
|
|
command: ["sleep", "infinity"]
|
|
# To run lammps, shell into the 0-0 pod:
|
|
# # Source environment
|
|
# . /mnt/flux/flux-view.sh
|
|
# # Connect to the running lead broker socket
|
|
# flux proxy $fluxsocket bash
|
|
# # See Flux resources!
|
|
# flux resource list
|
|
# Run lammps!
|
|
# flux run -N 4 -n 4 lmp -v x 2 -v y 2 -v z 2 -in in.reaxc.hns -nocite
|