GLM => nfolds = 20 family = gaussian solver = irlsm => nfolds getting set as 0 => wrong priority for task GLMSingleLambdaTsk, expected 0, but got 1

[exalate-issue-sync[bot]]

TestNG testcase : glm_neg_testcase_137

Test results page :

http://172.16.2.161:8080/view/testNG/job/h2o_master_DEV_testng_GLM_testcase/15/testngreports/h2o.testng/TestNG/glm_neg_testcase_137/

nfolds = 20 family = gaussian solver = irlsm

Validate Parameters object with testcase: glm_neg_testcase_137 Create modelParameter object with testcase: glm_neg_testcase_137 Set _family: gaussian Set _standardize: Set _lambda_search: Set _nfolds: 20 Set _ignore_const_cols: Set _non_negative: Set _intercept: Create train frame: airquality_train1 Create validate frame: airquality_train1 Set train frame Set validate frame Create success modelParameter object. Build model Train model 09-21 17:13:34.957 172.16.2.171:54321 24632 FJ-0-7 INFO: Creating 20 cross-validation splits with random number seed: -5596913177457903046 09-21 17:13:34.973 172.16.2.171:54321 24632 FJ-0-7 INFO: Building cross-validation model 1 / 20. 09-21 17:13:34.974 172.16.2.171:54321 24632 FJ-1-5 INFO: Building H2O GLM model with these parameters: 09-21 17:13:34.974 172.16.2.171:54321 24632 FJ-1-5 INFO: {"_model_id":null,"_train":{"name":"model_cv_1_airquality_train1.hex_train","type":"Key"},"_valid":{"name":"model_cv_1_airquality_train1.hex_valid","type":"Key"},"_nfolds":0,"_keep_cross_validation_predictions":false,"_fold_assignment":"AUTO","_distribution":"AUTO","_tweedie_power":1.5,"_ignored_columns":null,"_ignore_const_cols":false,"_weights_column":"weights","_offset_column":null,"_fold_column":null,"_score_each_iteration":false,"_response_column":"Ozone","_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,"_standardize":false,"_family":"gaussian","_link":"family_default","_solver":"IRLSM","_tweedie_variance_power":0.0,"_tweedie_link_power":1.0,"_alpha":null,"_lambda":null,"_prior":-1.0,"_lambda_search":false,"_nlambdas":100,"_non_negative":false,"_exactLambdas":false,"_lambda_min_ratio":-1.0,"_use_all_factor_levels":false,"_max_iterations":-1,"_intercept":false,"_beta_epsilon":1.0E-5,"_objective_epsilon":1.0E-5,"_gradient_epsilon":1.0E-4,"_beta_constraints":null,"_max_active_predictors":-1} java.lang.AssertionError: wrong priority for task GLMSingleLambdaTsk, expected 0, but got 1 at water.H2O$H2OCountedCompleter.compute(H2O.java:994) at jsr166y.CountedCompleter.exec(CountedCompleter.java:429) at jsr166y.ForkJoinTask.doExec(ForkJoinTask.java:263) at jsr166y.ForkJoinPool$WorkQueue.pollAndExecAll(ForkJoinPool.java:914) at jsr166y.ForkJoinPool$WorkQueue.runTask(ForkJoinPool.java:979) at jsr166y.ForkJoinPool.runWorker(ForkJoinPool.java:1477) at jsr166y.ForkJoinWorkerThread.run(ForkJoinWorkerThread.java:104) 09-21 17:13:34.977 172.16.2.171:54321 24632 FJ-1-5 INFO: GLM[dest=model_cv_1, iteration=0, lambda = 1877.9]: All 5 coefficients are active likelihood = 100648.0 09-21 17:13:34.981 172.16.2.171:54321 24632 FJ-0-13 WARN: ADMM solver reached maximum number of iterations (10000) 09-21 17:13:34.981 172.16.2.171:54321 24632 FJ-0-13 WARN: ADMM solver finished with gerr = 8449.675328571428 > eps = 1.0E-4 09-21 17:13:34.981 172.16.2.171:54321 24632 FJ-0-13 INFO: GLM[dest=model_cv_1, iteration=1, lambda = 1877.9]: iteration computed in 0 + 3 ms 09-21 17:13:34.981 172.16.2.171:54321 24632 FJ-0-13 INFO: GLM[dest=model_cv_1, iteration=1, lambda = 1877.9]: converged (reached a fixed point with ~ 1e-2147483648 precision), got 0 nzs 09-21 17:13:34.983 172.16.2.171:54321 24632 FJ-0-13 INFO: GLM[dest=model_cv_1, iteration=1, lambda = 1877.9]: hold-out set validation = mse = 109.0, explained_dev = 0.0 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: Solution at lambda = 1877.9142857142856 has 0 nonzeros, gradient err = 8449.675328571428 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: Model Metrics Type: RegressionGLM 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: Description: N/A 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: model id: model_cv_1 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: frame id: model_cv_1_airquality_train1.hex_train 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: MSE: 2875.6572 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: R^2: -1.8281012 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: mean residual deviance: 2875.6572 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: null DOF: 70.0 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: residual DOF: 70.0 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: null deviance: 201296.0 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: residual deviance: 201296.0 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: AIC: 758.134 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: Model Metrics Type: RegressionGLM 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: Description: N/A 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: model id: model_cv_1 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: frame id: model_cv_1_airquality_train1.hex_valid 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: MSE: 109.0 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: R^2: -11.111111 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: mean residual deviance: 109.0 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: null DOF: 2.0 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: residual DOF: 2.0 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: null deviance: 218.0 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: residual deviance: 218.0 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: AIC: 17.05845 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: GLM Model (summary): 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: Family Link Regularization Number of Predictors Total Number of Active Predictors Number of Iterations Training Frame 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: gaussian identity Elastic Net (alpha = 0.5, lambda = 1877.9 ) 6 1 1 model_cv_1_airquality_train1.hex_train 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: Scoring History: 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: timestamp duration iteration log_likelihood objective 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: 2015-09-21 17:13:34 0.000 sec 0 100648.00000 1437.82857 09-21 17:13:34.984 172.16.2.171:54321 24632 FJ-0-13 INFO: 2015-09-21 17:13:34 0.006 sec 1 100648.00000 1437.82857 09-21 17:13:34.985 172.16.2.171:54321 24632 FJ-0-7 INFO: Building cross-validation model 2 / 20. 09-21 17:13:34.986 172.16.2.171:54321 24632 FJ-1-5 INFO: Building H2O GLM model with these parameters: 09-21 17:13:34.986 172.16.2.171:54321 24632 FJ-1-5 INFO: {"_model_id":null,"_train":{"name":"model_cv_2_airquality_train1.hex_train","type":"Key"},"_valid":{"name":"model_cv_2_airquality_train1.hex_valid","type":"Key"},"_nfolds":0,"_keep_cross_validation_predictions":false,"_fold_assignment":"AUTO","_distribution":"AUTO","_tweedie_power":1.5,"_ignored_columns":null,"_ignore_const_cols":false,"_weights_column":"weights","_offset_column":null,"_fold_column":null,"_score_each_iteration":false,"_response_column":"Ozone","_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,"_standardize":false,"_family":"gaussian","_link":"family_default","_solver":"IRLSM","_tweedie_variance_power":0.0,"_tweedie_link_power":1.0,"_alpha":null,"_lambda":null,"_prior":-1.0,"_lambda_search":false,"_nlambdas":100,"_non_negative":false,"_exactLambdas":false,"_lambda_min_ratio":-1.0,"_use_all_factor_levels":false,"_max_iterations":-1,"_intercept":false,"_beta_epsilon":1.0E-5,"_objective_epsilon":1.0E-5,"_gradient_epsilon":1.0E-4,"_beta_constraints":null,"_max_active_predictors":-1} java.lang.AssertionError: wrong priority for task GLMSingleLambdaTsk, expected 0, but got 1 at water.H2O$H2OCountedCompleter.compute(H2O.java:994) at jsr166y.CountedCompleter.exec(CountedCompleter.java:429) at jsr166y.ForkJoinTask.doExec(ForkJoinTask.java:263) at jsr166y.ForkJoinPool$WorkQueue.pollAndExecAll(ForkJoinPool.java:914) at jsr166y.ForkJoinPool$WorkQueue.runTask(ForkJoinPool.java:979) at jsr166y.ForkJoinPool.runWorker(ForkJoinPool.java:1477) at jsr166y.ForkJoinWorkerThread.run(ForkJoinWorkerThread.java:104) 09-21 17:13:34.988 172.16.2.171:54321 24632 FJ-1-5 INFO: GLM[dest=model_cv_2, iteration=0, lambda = 1760.6]: All 5 coefficients are active likelihood = 90335.0

[exalate-issue-sync[bot]]

Tom Kraljevic commented: Tomas says this is fixed because glm now behaves like the other algos and doesn't have special priority issues.

[DinukaH2O]

JIRA Issue Migration Info

Jira Issue: PUBDEV-2127 Assignee: Tomas Nykodym Reporter: Neeraja Madabhushi State: Resolved Fix Version: N/A Attachments: N/A Development PRs: N/A