Bad ast created by R after deeplearning predict. seems to be missing the key for a row slice operation

Description

~/h2o-dev/h2o-r/tests/testdir_algos/deeplearning/runit_NOPASS_deeplearning_iris_getModel.R
changed name to above

going to move the test to no-pass
don't know what started making the bad AST, but it doesn't seem like it's anqi's push

I'll file a jira with this info

-kevin

On 02/24/2015 08:56 PM, Kevin wrote:
> It starts in job 273 in the commit-only run
> here is the java_12_0.out.txt
> There's a bad AST in the java out? But the R code doesn't seem to report this bad AST. It's right after the Predict, so I think that jives with the badness clamed around predict by the R test..??
>
> I don't see what's wrong with the AST, it looks like it's extracting rows 0 to 9. .
>
> oh maybe the key it's slicing is missing? it's assigning to a key, but I don't see the key name that it's slicing? with the ([ ) ..there should be a key name after the ([
>
>
>
>
> http://mr-0xb1:8080/job/h2o_master_DEV_runit_small_commit_only/273/artifact/h2o-r/tests/results/java_12_0.out.txt
> 02-24 19:08:11.456 172.16.2.164:41024 2051 # Session INFO: Method: POST , URI: /3/Predictions.json/models/DeepLearningModel_93b132d139bce15b6786e4c109b24962/frames/iris.hex, route: /3/Predictions/models/(?<model>.)/frames/(?<frame>.), parms: {model=DeepLearningModel_93b132d139bce15b6786e4c109b24962, frame=iris.hex}
> 02-24 19:08:11.472 172.16.2.164:41024 2051 # Session INFO: Confusion Matrix:
> 02-24 19:08:11.472 172.16.2.164:41024 2051 # Session INFO: Act/Pred Iris-setosa Iris-versicolor Iris-virginica Error
> 02-24 19:08:11.472 172.16.2.164:41024 2051 # Session INFO: Iris-setosa 50 0 0 0.0000 = 0 / 50
> 02-24 19:08:11.472 172.16.2.164:41024 2051 # Session INFO: Iris-versicolor 0 45 5 0.1000 = 5 / 50
> 02-24 19:08:11.472 172.16.2.164:41024 2051 # Session INFO: Iris-virginica 0 0 50 0.0000 = 0 / 50
> 02-24 19:08:11.472 172.16.2.164:41024 2051 # Session INFO: Totals 50 45 55 0.0333 = 10 / 300
> 02-24 19:08:11.510 172.16.2.164:41024 2051 # Session INFO: Method: DELETE, URI: /3/Remove.json, route: /3/Remove, parms: {key=subset_2_sid_ad48090e8c532a6d7349724633b245fb}
> 02-24 19:08:11.525 172.16.2.164:41024 2051 # Session INFO: Method: DELETE, URI: /3/Remove.json, route: /3/Remove, parms: {key=predictions_DeepLearningModel__93b132d139bce15b6786e4c109b24962_on_iris.hex}
> 02-24 19:08:11.550 172.16.2.164:41024 2051 # Session INFO: Method: GET , URI: /3/Rapids.json/isEval, route: /3/Rapids/isEval, parms: {ast_key=predictions_DeepLearningModel__93b132d139bce15b6786e4c109b24962_on_iris.hex}
> 02-24 19:08:11.574 172.16.2.164:41024 2051 # Session INFO: Method: GET , URI: /3/Rapids.json/isEval, route: /3/Rapids/isEval, parms: {ast_key=subset_3_sid_ad48090e8c532a6d7349724633b245fb}
>
> 02-24 19:08:11.601 172.16.2.164:41024 2051 # Session INFO: Method: POST , URI: /3/Rapids.json, route: /3/Rapids, parms: {ast=(= !subset_3_sid_ad48090e8c532a6d7349724633b245fb ([ (: #0 #9) "null"))}
> java.lang.IllegalArgumentException: End of input unexpected. Badly formed AST.
> at water.rapids.Exec.parse(Exec.java:108)
> at water.rapids.ASTSlice.parse_impl(AST.java:1030)
> at water.rapids.ASTSlice.parse_impl(AST.java:1020)
> at water.rapids.Exec.parse(Exec.java:113)
> at water.rapids.ASTAssign.parse_impl(AST.java:626)
> at water.rapids.ASTAssign.parse_impl(AST.java:616)
> at water.rapids.Exec.parse(Exec.java:113)
> at water.rapids.Exec.exec(Exec.java:70)
> at water.api.RapidsHandler.exec(RapidsHandler.java:36)
> at sun.reflect.GeneratedMethodAccessor7.invoke(Unknown Source)
> at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
> at java.lang.reflect.Method.invoke(Method.java:606)
> at water.api.Handler.handle(Handler.java:57)
> at water.api.RequestServer.handle(RequestServer.java:602)
> at water.api.RequestServer.serve(RequestServer.java:560)
> at water.NanoHTTPD$HTTPSession.run(NanoHTTPD.java:433)
> at java.lang.Thread.run(Thread.java:745)
>
>
>
>
> seems to be from anqi's big push here at 7:04 PM, but I don't see anything she might have changed
> anqi didn't change any R to h2o stuff?
>
>
> arno did some pushes around 4:34 and 4:53 but there were multiple greens on those so unlikely those were new intermittents.
>
> [2015-02-24 19:46:58] [ERROR] : Error: Test failed: 'Deep Learning Test: Iris'
> Not expected: error in evaluating the argument 'x' in selecting a method for function 'print': Failed lookup of variable: predictions_DeepLearningModel__a9a5dc40b87eb1062980b7deed0d4eaf_on_iris.hex
>
> http://mr-0xb1:8080/job/h2o_master_DEV_runit_small_commit_only/273/
>
> Started 1 hr 29 min ago
> Took 4 min 26 sec on mr-0xb4
> Failed Build #273 (Feb 24, 2015 7:04:53 PM)
> Build Artifacts
> Changes
>
> WIP Implementing generalized low rank model for computing principal components in parallel (commit: 593dec3c8ed00c3ec37fa58f215ba4fc175f1e91) (detail / githubweb)
> WIP Implemented distributed solve step to obtain Y matrix (commit: d44b1cd6498e971e1d5457ae20ec47b35dfabdfe) (detail / githubweb)
> WIP Finished single optimization step of GLRM. (commit: 8ebaf968676d3290e2dddf46b52bbb6a30248f55) (detail / githubweb)
> Completed main iteration and consolidated a few MRTasks for efficiency. Initial Y matrix chosen by kmeans++. (commit: 9b6d385f77c8438864e1c93e7d5cfb377af9e26d) (detail / githubweb)
> WIP Improved efficiency of algorithm and fixed a few runtime bugs (commit: 129e2ddd19e3ad7346e64f0cffac82a555047325) (detail / githubweb)
> Make a few general math functions public static and fix an AIOOBE in SMulTask (commit: 702853cdad053a441268d09e69d0756427a48d4e) (detail / githubweb)
> Fixing some leaked key issues and a bug where X values were not updating properly. (commit: 430ee8477a0b2dc095773c18f7541efabd0d6b6c) (detail / githubweb)
> Fix some leaked key issues (commit: 18f8693afd5b89bf87030d83aefe1e592ee02b3e) (detail / githubweb)
> Fix some leaked key issues (commit: 76859741d58b46a689b1cb809a42689d214a3f76) (detail / githubweb)
> Convert to using H2O Cholesky method. Enable standardization of training data in MRTask. (commit: 593521ad67596b15674511e7bcbbd332d0d6d5e1) (detail / githubweb)
> WIP Recover principal components from GLRM results and continue testing for accuracy (commit: 23e7c235457b511c0dc5c1435bc4b513649e61a5) (detail / githubweb)
> Fixed some more bugs and reverted back to using Jama Cholesky (commit: 1d6b48f3355f0efd64afac211cd7767a9bf65306) (detail / githubweb)
> Added ability for user initialized Y matrix. Fix PCA recovery function so standard deviations are correctly computed. (commit: c4d788a7619f8441d96b64d7958e3df5f2be0fec) (detail / githubweb)
> Junit is finally passing with both demeaned and standardized data compared to R PCA (commit: 882e5d402d2703903412c1a7a5ddcf6641c3f652) (detail / githubweb)
> Fix a typo in Y matrix initialization (commit: ab90cc18ebce08864b3820aa6d85261b2405747a) (detail / githubweb)
> Add regularization to Gram matrix until SPD so Cholesky works (commit: b5170ba79ad5efe1b1bc06bea2fba661cf72c083) (detail / githubweb)
> WIP Adding GLRM to R package (commit: 4e18c6d5c348a0eea03c208755b76b81d203fe72) (detail / githubweb)
> WIP Connect GLRM to API. Throw non SPD exception if Cholesky still not SPD after multiple regularization attempts (commit: 6e67817bf5c04c0b6b669818c48e2a0e50b9ed65) (detail / githubweb)
> Update GLRM API model output (commit: 653a9906003c2bab5e65d576420c76093fc92d12) (detail / githubweb)
> Fix a bug preventing multiple iterations of alternating minimization. Update R call to GLRM so comparable to built in PCA. (commit: 62030056f08d3287c60ac811c5fc9018c2c00345) (detail / githubweb)
> Fix a AIOOBE in array multiplication. Add golden tests for GLRM comparing with R's PCA (commit: c5215c99328e5164ca3408e918589d5ee7a1eb09) (detail / githubweb)
> Fix a leaked key bug and incorrect calculation of objective function (commit: 61c56e5af0103a792ffc5bf2e0de98bf19f2096e) (detail / githubweb)
> Merge and update schema with master (commit: 63c9a52c40621d9b7b9cb677f9788cb847e339dc) (detail / githubweb)
> Massive cleanup and removal of old PCA, replacing with quadratically regularized PCA based on alternating minimization algorithm in GLRM (commit: 02b7f168b2efa551a60c4bf2e95b8d506b613c2d) (detail / githubweb)
> Add R demos back in for PCA and update golden tests (commit: 9855af94df86946101dd5f17273abdff50163a93) (detail / githubweb)
> Fix a bug introduced by incorrect auto merge. Update REST API test to check for PCA algorithm. (commit: 6f0aad3475e7c95888d517a36ffae919e263c2ca) (detail / githubweb)
>
>
>
>
>
>
> On 02/24/2015 08:28 PM, Neeraja Madabhushi wrote:
>> Hello All,
>> RUnit builds failing for more than 20 builds from 7:08 PM.......
>> ====================================
>> runit_deeplearning_iris_getModel.R
>> ===================================
>>
>> First 10 rows:
>>
>> ######## ### #### ##
>> ## ## ## ## ##
>> ## ## ## ## ##
>> ###### ## ## ## ##
>> ## ######### ## ##
>> ## ## ## ## ##
>> ## ## ## #### ########
>>
>> [2015-02-24 19:46:58] [ERROR] : Error: Test failed: 'Deep Learning Test: Iris'
>> Not expected: error in evaluating the argument 'x' in selecting a method for function 'print': Failed lookup of variable: predictions_DeepLearningModel__a9a5dc40b87eb1062980b7deed0d4eaf_on_iris.hex
>>
>> 1: withWarnings(test(conn))
>> 2: withCallingHandlers(expr, warning = wHandler)
>> 3: test(conn)
>> 4: print(predict(m, iris.hex))
>> 5: print(predict(m, iris.hex))
>> 6: print.default(predict(m, iris.hex))
>> 7: structure(function (object)
>> standardGeneric("show"), generic = structure("show", package = "methods"), package = "methods", group = list(), valueClass = character(0), signature = structure("object", simpleOnly = TRUE), default = structure(function (object)
>> showDefault(object, FALSE), target = structure("ANY", class = structure("signature", package = "methods"), .Names = "object", package = "methods"), defined = structure("ANY", class = structure("signature", package = "methods"), .Names = "object", package = "methods"), generic = structure("show", package = "methods"), class = structure("derivedDefaultMethod", package = "methods")), skeleton = (structure(function (object)
>> showDefault(object, FALSE), target = structure("ANY", class = structure("signature", package = "methods"), .Names = "object", package = "methods"), defined = structure("ANY", class = structure("signature", package = "methods"), .Names = "object", package = "methods"), generic = structure("show", package = "methods"), class = structure("derivedDefaultMethod", package = "methods")))(object), class = structure("standardGeneric", package = "methods"))(<S4 object of class structure("H2OFrame", package = "h2o")>)
>> 8: structure(function (object)
>> standardGeneric("show"), generic = structure("show", package = "methods"), package = "methods", group = list(), valueClass = character(0), signature = structure("object", simpleOnly = TRUE), default = structure(function (object)
>> showDefault(object, FALSE), target = structure("ANY", class = structure("signature", package = "methods"), .Names = "object", package = "methods"), defined = structure("ANY", class = structure("signature", package = "methods"), .Names = "object", package = "methods"), generic = structure("show", package = "methods"), class = structure("derivedDefaultMethod", package = "methods")), skeleton = (structure(function (object)
>> showDefault(object, FALSE), target = structure("ANY", class = structure("signature", package = "methods"), .Names = "object", package = "methods"), defined = structure("ANY", class = structure("signature", package = "methods"), .Names = "object", package = "methods"), generic = structure("show", package = "methods"), class = structure("derivedDefaultMethod", package = "methods")))(object), class = structure("standardGeneric", package = "methods"))(<S4 object of class structure("H2OFrame", package = "h2o")>)
>> 9: print(head(object, 10L))
>> 10: .handleSimpleError(function (e)
>> {
>> e$calls <- head(sys.calls()[-seq_len(frame + 7)], -2)
>> signalCondition(e)
>> }, "error in evaluating the argument 'x' in selecting a method for function 'print': Error: Failed lookup of variable: predictions_DeepLearningModel__a9a5dc40b87eb1062980b7deed0d4eaf_on_iris.hex\n",
>> quote(print(head(object, 10L)))).
>>
>> SEED used: 686883847
>>
>> [2015-02-24 19:46:58] [ERROR] : TEST FAILED
>> No traceback available
>> ==============
>> Thanks,
>> Neeraja
>

Activity

Show:
Spencer Aiello
March 7, 2016, 7:10 PM

fixed in ancient history

Fixed
Your pinned fields
Click on the next to a field label to start pinning.

Assignee

Spencer Aiello

Reporter

Kevin Normoyle

CustomerVisible

No