systemml-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Fei Hu (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (SYSTEMML-1774) Improve Parfor parallelism for deep learning
Date Mon, 17 Jul 2017 18:52:01 GMT

    [ https://issues.apache.org/jira/browse/SYSTEMML-1774?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16090325#comment-16090325
] 

Fei Hu commented on SYSTEMML-1774:
----------------------------------

The errors when running the [distributed MNIST LeNet example] |  https://github.com/apache/systemml/blob/master/scripts/nn/examples/mnist_lenet_distrib_sgd.dml]
at the local machine:

{code:java}
Exception in thread "main" org.apache.sysml.api.mlcontext.MLContextException: Exception when
executing script
	at org.apache.sysml.api.mlcontext.MLContext.execute(MLContext.java:320)
	at org.apache.sysml.api.mlcontext.MLContext.execute(MLContext.java:293)
	at org.apache.sysml.api.mlcontext.Script.execute(Script.java:717)
	at org.apache.sysml.scripts.nn.examples.Mnist_lenet_distrib_sgd.train(Mnist_lenet_distrib_sgd.java)
	at org.apache.sysml.examples.MNIST_Distrib_Sgd$.main(MNIST_Distrib_Sgd.scala:41)
	at org.apache.sysml.examples.MNIST_Distrib_Sgd.main(MNIST_Distrib_Sgd.scala)
Caused by: org.apache.sysml.api.mlcontext.MLContextException: Exception occurred while executing
runtime program
	at org.apache.sysml.api.mlcontext.ScriptExecutor.executeRuntimeProgram(ScriptExecutor.java:397)
	at org.apache.sysml.api.mlcontext.ScriptExecutor.execute(ScriptExecutor.java:311)
	at org.apache.sysml.api.mlcontext.MLContext.execute(MLContext.java:316)
	... 5 more
Caused by: org.apache.sysml.runtime.DMLRuntimeException: org.apache.sysml.runtime.DMLRuntimeException:
ERROR: Runtime error in program block generated from statement block between lines 11 and
11 -- Error evaluating instruction: CP°extfunct°./scripts/nn/examples/mnist_lenet_distrib_sgd.dml°train°10°8°X·MATRIX·DOUBLE°Y·MATRIX·DOUBLE°X_val·MATRIX·DOUBLE°Y_val·MATRIX·DOUBLE°C·SCALAR·INT·false°Hin·SCALAR·INT·false°Win·SCALAR·INT·false°batch_size·SCALAR·INT·false°parallel_batches·SCALAR·INT·false°epochs·SCALAR·INT·false°W1°b1°W2°b2°W3°b3°W4°b4
	at org.apache.sysml.runtime.controlprogram.Program.execute(Program.java:130)
	at org.apache.sysml.api.ScriptExecutorUtils.executeRuntimeProgram(ScriptExecutorUtils.java:95)
	at org.apache.sysml.api.ScriptExecutorUtils.executeRuntimeProgram(ScriptExecutorUtils.java:54)
	at org.apache.sysml.api.mlcontext.ScriptExecutor.executeRuntimeProgram(ScriptExecutor.java:395)
	... 7 more
Caused by: org.apache.sysml.runtime.DMLRuntimeException: ERROR: Runtime error in program block
generated from statement block between lines 11 and 11 -- Error evaluating instruction: CP°extfunct°./scripts/nn/examples/mnist_lenet_distrib_sgd.dml°train°10°8°X·MATRIX·DOUBLE°Y·MATRIX·DOUBLE°X_val·MATRIX·DOUBLE°Y_val·MATRIX·DOUBLE°C·SCALAR·INT·false°Hin·SCALAR·INT·false°Win·SCALAR·INT·false°batch_size·SCALAR·INT·false°parallel_batches·SCALAR·INT·false°epochs·SCALAR·INT·false°W1°b1°W2°b2°W3°b3°W4°b4
	at org.apache.sysml.runtime.controlprogram.ProgramBlock.executeSingleInstruction(ProgramBlock.java:316)
	at org.apache.sysml.runtime.controlprogram.ProgramBlock.executeInstructions(ProgramBlock.java:217)
	at org.apache.sysml.runtime.controlprogram.ProgramBlock.execute(ProgramBlock.java:163)
	at org.apache.sysml.runtime.controlprogram.Program.execute(Program.java:123)
	... 10 more
Caused by: org.apache.sysml.runtime.DMLRuntimeException: error executing function ./scripts/nn/examples/mnist_lenet_distrib_sgd.dml::train
	at org.apache.sysml.runtime.instructions.cp.FunctionCallCPInstruction.processInstruction(FunctionCallCPInstruction.java:187)
	at org.apache.sysml.runtime.controlprogram.ProgramBlock.executeSingleInstruction(ProgramBlock.java:286)
	... 13 more
Caused by: org.apache.sysml.runtime.DMLRuntimeException: ERROR: Runtime error in function
program block generated from function statement block between lines 36 and 271 -- Error evaluating
function program block
	at org.apache.sysml.runtime.controlprogram.FunctionProgramBlock.execute(FunctionProgramBlock.java:122)
	at org.apache.sysml.runtime.instructions.cp.FunctionCallCPInstruction.processInstruction(FunctionCallCPInstruction.java:180)
	... 14 more
Caused by: org.apache.sysml.runtime.DMLRuntimeException: ERROR: Runtime error in for program
block generated from for statement block between lines 115 and 270 -- Error evaluating for
program block
	at org.apache.sysml.runtime.controlprogram.ForProgramBlock.execute(ForProgramBlock.java:157)
	at org.apache.sysml.runtime.controlprogram.FunctionProgramBlock.execute(FunctionProgramBlock.java:115)
	... 15 more
Caused by: org.apache.sysml.runtime.DMLRuntimeException: ERROR: Runtime error in for program
block generated from for statement block between lines 117 and 265 -- Error evaluating for
program block
	at org.apache.sysml.runtime.controlprogram.ForProgramBlock.execute(ForProgramBlock.java:157)
	at org.apache.sysml.runtime.controlprogram.ForProgramBlock.execute(ForProgramBlock.java:145)
	... 16 more
Caused by: org.apache.sysml.runtime.DMLRuntimeException: PARFOR: Failed to execute loop in
parallel.
	at org.apache.sysml.runtime.controlprogram.ParForProgramBlock.execute(ParForProgramBlock.java:660)
	at org.apache.sysml.runtime.controlprogram.ForProgramBlock.execute(ForProgramBlock.java:145)
	... 17 more
Caused by: org.apache.sysml.runtime.DMLRuntimeException: Not supported: Instructions of type
other than CP instructions org.apache.sysml.runtime.instructions.spark.RandSPInstruction
SPARK°rand°¶_Var3092¶°512°1000°1000°1°1°0.5°-1°scratch_space/_p10640_9.108.91.12//_t0/°uniform°1.0°_mVar3093·MATRIX·DOUBLE
	at org.apache.sysml.runtime.controlprogram.parfor.ProgramConverter.serializeInstructions(ProgramConverter.java:937)
	at org.apache.sysml.runtime.controlprogram.parfor.ProgramConverter.rSerializeProgramBlock(ProgramConverter.java:1297)
	at org.apache.sysml.runtime.controlprogram.parfor.ProgramConverter.rSerializeProgramBlocks(ProgramConverter.java:1125)
	at org.apache.sysml.runtime.controlprogram.parfor.ProgramConverter.rSerializeProgramBlock(ProgramConverter.java:1264)
	at org.apache.sysml.runtime.controlprogram.parfor.ProgramConverter.rSerializeFunctionProgramBlocks(ProgramConverter.java:1106)
	at org.apache.sysml.runtime.controlprogram.parfor.ProgramConverter.serializeProgram(ProgramConverter.java:805)
	at org.apache.sysml.runtime.controlprogram.parfor.ProgramConverter.serializeParForBody(ProgramConverter.java:763)
	at org.apache.sysml.runtime.controlprogram.ParForProgramBlock.executeRemoteSparkParFor(ParForProgramBlock.java:1034)
	at org.apache.sysml.runtime.controlprogram.ParForProgramBlock.execute(ParForProgramBlock.java:647)
	... 18 more
{code}


> Improve Parfor parallelism for deep learning
> --------------------------------------------
>
>                 Key: SYSTEMML-1774
>                 URL: https://issues.apache.org/jira/browse/SYSTEMML-1774
>             Project: SystemML
>          Issue Type: Improvement
>          Components: Algorithms, Compiler, ParFor
>    Affects Versions: SystemML 1.0
>            Reporter: Fei Hu
>              Labels: deeplearning
>
> When running the  [distributed MNIST LeNet example | https://github.com/apache/systemml/blob/master/scripts/nn/examples/mnist_lenet_distrib_sgd.dml],
each mini-batch could ideally run in parallel without interaction. We try to force {{parfor
(j in 1:parallel_batches)}} at line 137 of {{nn/examples/mnist_lenet_distrib_sgd.dml}} to
be {{parfor (j in 1:parallel_batches, mode=REMOTE_SPARK, opt=CONSTRAINED)}} use {{REMOTE_SPARK}}
mode, but got some errors about {{org.apache.sysml.runtime.DMLRuntimeException: Not supported:
Instructions of type other than CP instructions}}. More log information can be found at the
following comments. One example of the errors is that at the convolution layer, we need to
randomly generate a matrix, but SystemML choose {{RandSPInstruction}} instead of {{DataGenCPInstruction}},
which may be because SystemML could not determine the row number of the matrix. For this distributed
MNIST LeNet  example, using CPInstruction may achieve better performance. 



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

Mime
View raw message