Skip to content

Latest commit

 

History

History
49 lines (46 loc) · 4.55 KB

submit.md

File metadata and controls

49 lines (46 loc) · 4.55 KB

Submit Parameter

中文文档

Using the command $HBOX_HOME/bin/hbox-submit [[--property value]...] command [argument...] to submit the application to Cluster at the Hbox client. Please see the example in the part of README Quick Start. The following is more details of the parameter.

Property Name Meaning
app-name application name
app-type application type, default as the "Hbox", can set as "TensorFlow", "Caffe" according to the deeplearning framework
input input file path in the format of "the HDFS path"#"local path"
output output file path in the format of "the HDFS path"#"local path"
files the required local files of the application
cacheArchive the required compressed files in the HDFS path
cacheFile the required files in the HDFS path
user-path the append for the environment variable $PATH
jars the required jar files
user-classpath-first whether user job jar should be the first one on class path or not, default as the configure of hbox.user.classpath.first
conf set the configuration
driver-cores number of cores to use for the AM process, default as the configure of hbox.driver.cores
driver-memory amount of memory to use for the AM process (in MB),default as the configure of hbox.driver.memory
ps-num number of ps containers to use for the application, default as the configure of hbox.ps.num
ps-cores number of cores to use for the ps process, default as the configure of hbox.ps.cores
ps-memory amount of memory to use for the ps process (in MB), default as the configure of hbox.ps.memory
worker-num number of worker containers to use for the application, default as the configure of hbox.worker.num
worker-cores number of cores to use for the worker process, default as the configure of hbox.worker.cores
worker-memory amount of memory to use for the worker process(in MB), default as the configure of hbox.worker.memory
chiefworker-memory amount of memory for the chief worker, especially for the index 0 worker of the TensorFlow application, default as the worker-memory
evaluatorworker-memory amount of memory for the estimator worker, especially for the TensorFlow Estimator application, default as the worker-memory
queue the queue of application submitted to, default as the configure of hbox.app.queue
priority the priority of application, default as the configure of hbox.app.priority
board-enable whether to start the service of Board, default as the configure of hbox.tf.board.enable
board-index specify the index of worker which start the Board, default as the configure of hbox.tf.board.worker.index
board-logdir the directory save Board event log, default as the configure of hbox.tf.board.log.dir
board-reloadinterval how often the backend should load more data of event log for tensorboard, default as the configure of hbox.tf.board.reload.interval
board-historydir specify the HDFS path which the Board event log upload to, default as the configure of hbox.tf.board.history.dir
board-modelpb model proto in ONNX format for VisualDL, default as the configure of hbox.board.modelpb
board-cacheTimeout memory cache timeout duration in seconds for VisualDL,default as the configure of hbox.board.cache.timeout
input-strategy the strategy of the input file, default as the configure of hbox.input.strategy
inRenameInputFile whether to rename the download file when input-strategy is "DOWNLOAD", default as the configure of hbox.inputfile.rename
stream-epoch specify the epoch num of the input file read when input-strategy is "STREAM", default as the configure of hbox.stream.epoch
inputformat specify the class of the inputformat when input-strategy is "STREAM", default as the configure of hbox.inputformat.class
inputformat-shuffle whether to shuffle the input splits when input-strategy is "STREAM", default as the configure of hbox.input.stream.shuffle
output-strategy the strategy of the output file, default as the configure of hbox.output.strategy
outputformat specify the class of outputformat when output-strategy is "STREAM", default as the configure of hbox.outputformat.class
tf-evaluator whether to set the last worker as evaluator of distributed TensorFlow job type, default as the configure of hbox.tf.evaluator
output-index specify the index of the worker which to upload the output, default upload the output of all the workers.
archiveFiles Location of local archive files will be uploaded to container and be decompressed. use comma as separator, # with alias name