Hendrik/xgboost
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

yarn script

Browse Source
This commit is contained in:
chenshuaihua 2015-01-11 23:19:04 +08:00
parent d81fb6a9e6
commit b5ac85f103
1 changed files with 10 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
tracker/rabit_yarn.py
View File

@ -46,7 +46,14 @@ parser.add_argument('-f', '--files', nargs = '*',
' the submission script will automatically cache all the files which appears in command.'\
' You may need this option to cache additional files.'\
' You can also use it to manually cache files when auto_file_cache is off')
parser.add_argument('--jobname', help = 'customize jobname in tracker')
parser.add_argument('--jobname', default='auto', help = 'customize jobname in tracker')
parser.add_argument('--timeout', default=600000000, type=int,
help = 'timeout (in million seconds) of each mapper job, automatically set to a very long time,'\
'normally you do not need to set this ')
parser.add_argument('-m', '--memory_mb', default=1024, type=int,
help = 'maximum memory used by the process, Guide: set it large (near mapreduce.jobtracker.maxmapmemory.mb).'\
'if you are running multi-threading rabit,'\
'so that each node can occupy all the mapper slots in a machine for maximum performance')
if hadoop_binary == None:
parser.add_argument('-hb', '--hadoop_binary', required = True,
help="path-to-hadoop binary folder")
@ -71,6 +78,8 @@ def hadoop_streaming(nworker, slave_args):
cmd = '%s jar %s -D mapreduce.job.maps=%d' % (args.hadoop_binary, args.hadoop_streaming_jar, nworker)
cmd += ' -D mapreduce.job.name=%s' % (args.jobname)
cmd += ' -D mapreduce.map.cpu.vcores=%d' % (args.nthread)
cmd += ' -D mapreduce.task.timeout=%d' % (args.timeout)
cmd += ' -D mapreduce.map.memory.mb=%d' % (args.memory_mb)
cmd += ' -input %s -output %s' % (args.input, args.output)
cmd += ' -mapper \"%s\" -reducer \"/bin/cat\" ' % (' '.join(args.command + slave_args))
fset = set()