Unable to modify data in H2O frame.

[bloog...@gmail.com]

I'm trying to adjust weights used in my panel regression.
But I get a 'Can't unlock: Not locked!' error.

Are there other ways to modify data in H2O?

-Brian




    dts = as.data.frame(curDatH$Date)                                           ### h2o ->R
    tsWts = cal.bdayDiff( dts$Date, useUpTo, cal)
    tsWts = 2 ^ -( tsWts / 252 )                                                      ### Calc exponential weights
    tsWtsH = as.h2o( localH20, tsWts, destination_frame= "wts")   


    curDatH$Wts <- curDatH$Wts * tsWtsH                                    ### Adjust weights in h2o


Got exception 'class java.lang.AssertionError', with msg 'Can't unlock: Not locked!'
java.lang.AssertionError: Can't unlock: Not locked!
        at water.Lockable$Unlock.atomic(Lockable.java:181)
        at water.Lockable$Unlock.atomic(Lockable.java:176)
        at water.TAtomic.atomic(TAtomic.java:17)
        at water.Atomic.compute2(Atomic.java:55)
        at water.Atomic.fork(Atomic.java:39)
        at water.Atomic.invoke(Atomic.java:31)
        at water.Lockable.unlock(Lockable.java:171)
        at hex.Model$Parameters.read_unlock_frames(Model.java:170)
        at hex.tree.SharedTree$Driver.compute2(SharedTree.java:231)
        at water.H2O$H2OCountedCompleter.compute(H2O.java:1002)
        at jsr166y.CountedCompleter.exec(CountedCompleter.java:429)
        at jsr166y.ForkJoinTask.doExec(ForkJoinTask.java:263)
        at jsr166y.ForkJoinPool$WorkQueue.runTask(ForkJoinPool.java:974)
        at jsr166y.ForkJoinPool.runWorker(ForkJoinPool.java:1477)
        at jsr166y.ForkJoinWorkerThread.run(ForkJoinWorkerThread.java:104)





### if it worked I would then fit the model

    fitH = h2o.gbm(y = dependent, x = independent, weights_column = weights,  training_frame = curDatH,
                ntrees = 100, max_depth = 10, min_rows = 200, nfolds=3, learn_rate = 0.1,
                modelID = paste("GBM",fitDt,sep="_") )

    summary(fitH)

[bloog...@gmail.com]

Still confused on this...

[Spencer Aiello]

Hi Brian,

Do happen to have the logs available? Or would it be alright to share a snippet of your data so that we may attempt a reproduction? You may pm it to me if you like (spe...@h2o.ai).

Spencer

--
You received this message because you are subscribed to the Google Groups "H2O Open Source Scalable Machine Learning - h2ostream" group.
To unsubscribe from this group and stop receiving emails from it, send an email to h2ostream+...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[bloog...@gmail.com]

Hi Spencer,
I found the logs stored in /tmp/

Interestingly it seems that it was complaining on 'Can't unlock: Not locked!'

the logs show something different:
#######h2o_127.0.0.1_54321-3-info.log"
10-14 22:35:40.818 127.0.0.1:54321       32041  FJ-0-39   ERRR: _min_rows: The dataset size is too small to split for min_rows=10.0: must have at least 20.0 (weighted) rows, but have only 19.38260765603209.

#######h2o_127.0.0.1_54321-2-debug.log
10-14 22:35:40.817 127.0.0.1:54321       32041  FJ-0-39   INFO: Building H2O GBM model with these parameters:
10-14 22:35:40.817 127.0.0.1:54321       32041  FJ-0-39   INFO: {"_model_id":null,"_train":{"name":"GBM_model_R_1444875645239_55_cv_1_subset_5422_train","type":"Key"},"_valid":{"name":"GBM_model_R_1444875645239_55_cv_1_subset_5422_valid","type":"Key"},"_nfolds":0,"_keep_cross_validation_predictions":false,"_fold_assignment":"AUTO","_distribution":"AUTO","_tweedie_power":1.5,"_ignored_columns":["Date","ID","Int","Ido"],"_ignore_const_cols":true,"_weights_column":"weights","_offset_column":null,"_fold_column":null,"_score_each_iteration":false,"_response_column":"sumRet","_balance_classes":false,"_max_after_balance_size":5.0,"_class_sampling_factors":null,"_max_hit_ratio_k":10,"_max_confusion_matrix_size":20,"_checkpoint":null,"_ntrees":100,"_max_depth":10,"_min_rows":10.0,"_nbins":20,"_nbins_cats":1024,"_r2_stopping":0.999999,"_seed":-2484894243866892320,"_nbins_top_level":1024,"_build_tree_one_node":false,"_initial_score_interval":4000,"_score_interval":4000,"_learn_rate":0.1}
10-14 22:35:40.818 127.0.0.1:54321       32041  FJ-0-39   INFO: Dropping ignored columns: [Date, ID, Intercept, Idio]
10-14 22:35:40.818 127.0.0.1:54321       32041  FJ-0-39   ERRR: _min_rows: The dataset size is too small to split for min_rows=10.0: must have at least 20.0 (weighted) rows, but have only 19.38260765603209.
10-14 22:35:40.822 127.0.0.1:54321       32041  FJ-0-39   DEBUG: unlock GBM_model_R_1444875645239_55_cv_1_subset_5422_train by job $03017f00000132d4ffffffff$_a24300cbb0eb4b05e64a9fd6b5e046bb_cv0


I've removed the min row count and

To unsubscribe from this group and stop receiving emails from it, send an email to h2ostream+unsubscribe@googlegroups.com.

[bloog...@gmail.com]

I've removed the min row count threshold and it seems to work fine now.
Dont think it was ever any locking type of problem.

On Thursday, October 15, 2015 at 5:40:19 PM UTC-4, bloog...@gmail.com wrote:

Hi Spencer,
I found the logs stored in /tmp/

Interestingly it seems that it was complaining on 'Can't unlock: Not locked!'

the logs show something different:
#######h2o_127.0.0.1_54321-3-info.log"
10-14 22:35:40.818 127.0.0.1:54321       32041  FJ-0-39   ERRR: _min_rows: The dataset size is too small to split for min_rows=10.0: must have at least 20.0 (weighted) rows, but have only 19.38260765603209.

#######h2o_127.0.0.1_54321-2-debug.log
10-14 22:35:40.817 127.0.0.1:54321       32041  FJ-0-39   INFO: Building H2O GBM model with these parameters:
10-14 22:35:40.817 127.0.0.1:54321       32041  FJ-0-39   INFO: {"_model_id":null,"_train":{"name":"GBM_model_R_1444875645239_55_cv_1_subset_5422_train","type":"Key"},"_valid":{"name":"GBM_model_R_1444875645239_55_cv_1_subset_5422_valid","type":"Key"},"_nfolds":0,"_keep_cross_validation_predictions":false,"_fold_assignment":"AUTO","_distribution":"AUTO","_tweedie_power":1.5,"_ignored_columns":["Date","ID","Int","Ido"],"_ignore_const_cols":true,"_weights_column":"weights","_offset_column":null,"_fold_column":null,"_score_each_iteration":false,"_response_column":"sumRet","_balance_classes":false,"_max_after_balance_size":5.0,"_class_sampling_factors":null,"_max_hit_ratio_k":10,"_max_confusion_matrix_size":20,"_checkpoint":null,"_ntrees":100,"_max_depth":10,"_min_rows":10.0,"_nbins":20,"_nbins_cats":1024,"_r2_stopping":0.999999,"_seed":-2484894243866892320,"_nbins_top_level":1024,"_build_tree_one_node":false,"_initial_score_interval":4000,"_score_interval":4000,"_learn_rate":0.1}
10-14 22:35:40.818 127.0.0.1:54321       32041  FJ-0-39   INFO: Dropping ignored columns: [Date, ID, Intercept, Idio]
10-14 22:35:40.818 127.0.0.1:54321       32041  FJ-0-39   ERRR: _min_rows: The dataset size is too small to split for min_rows=10.0: must have at least 20.0 (weighted) rows, but have only 19.38260765603209.
10-14 22:35:40.822 127.0.0.1:54321       32041  FJ-0-39   DEBUG: unlock GBM_model_R_1444875645239_55_cv_1_subset_5422_train by job $03017f00000132d4ffffffff$_a24300cbb0eb4b05e64a9fd6b5e046bb_cv0

I've removed the min row count threshold and it seems to work fine now.