in dags/map_reproducibility/utils/sample_workload_utils.py [0:0]
def sample_workload_gcs_to_cns_cmds(log_file_in_gcs, output_file=None):
# This function only works for glinux or cloudtop because it is using fileutil_bs
# If output_file is not provided, use the same name as the input file
log_file_in_gcs = log_file_in_gcs.removeprefix("gs://")
if not output_file:
output_file = os.path.basename(log_file_in_gcs)
print(f"output_file name is: {output_file}")
cmds = (
f"LOG_FILE_IN_GCS={log_file_in_gcs} ",
f"filename={output_file} ",
"CNS_PATH=/cns/pi-d/home/${USER}/tensorboard/multislice ",
"/google/data/ro/projects/cloud/bigstore/mpm/fileutil_bs/stable/bin/fileutil_bs cp /bigstore/${LOG_FILE_IN_GCS} ${CNS_PATH}/${filename} ",
"echo file to put into xprof: ${CNS_PATH}/${filename}",
)
return cmds