clean up example

Signed-off-by: vsoch <[email protected]>

Author: vsoch

examples/test/jobset-mnist.yaml

@@ -7,18 +7,6 @@ spec:
     pullPolicy: Never
   workflow:
     completed: 10
-     # What we could try:
-     # Having a change to epochs
-     # Stopping the workflow when we reach a certain threshold accuracy, etc.
-#    events:
-       # Custom metric - derived from parsing lammps log
-        # Max decreases down to 1 (default), with 3 breaks in between, and 3 times
-#      - metric: mean.lammps.lammps-walltime
-#        when: "<= 10"
-#        action: shrink
-#        repetitions: 3
-#        backoff: 3
-#        minSize: 1
 
   cluster:
     maxSize: 4
@@ -40,14 +28,3 @@ spec:
       PYTHONUNBUFFERED: "0"
       epochs: "1"
     script: torchrun --rdzv_id=123 --nnodes=${nodes} --nproc_per_node=1 --master_addr=${jobname}-jobset-0-0.${jobname}.default.svc.cluster.local --master_port=$MASTER_PORT --node_rank=$RANK mnist.py --epochs=${epochs} --log-interval=1
-
-    # Event parsing. Assume for a log for now
-#    events:
-#     script: |
-#        def parse_log(log):
-#            import re
-#            match = re.search('Total wall time: (?P<walltime>.*)', log)
-#            walltime = match.groupdict()['walltime']
-#            hours, minutes, seconds = walltime.split(':')
-#            walltime = (float(hours) * 60 * 60) + (float(minutes) * 60) + (float(seconds))
-#            return {"lammps-walltime": walltime}