Public H2O 2 [OLD]
Back to project
Issues and filters
View all filters
Public H2O 2 [OLD]
Unable to create subset based on between values by referring first index
Stringdist (call from R) throws an error when frames are larger than 1000 records
a phantomjs test fail suggests parsing from s3 might have a memory leak
water meter in flow has a memory leak which eventually makes h2o crash with out of memory error
Parsing issue and inconsistencies
h2oR drf: if sample ratio is 1 and no validation set provided, get Error in 1 - result$accuracy : non-numeric argument to binary operator
Detect and report co-hosted nodes
GLM: when run with offset get java.lang.AssertionError
Increase GA reporting
runit_NOPASS_quantile_1_golden.R: intermittent hang
runit_NOPASS_demo_glm_uuid.R: intermittent hang on quantile...
Output detected MTU to logs
Output Hadoop information to logs
All issues related to useability in launching H2O, clarity in failure feedback, and tools and instrumentation for faster debugging
Hadoop drivers should detect and report when being run on the wrong version of Hadoop
runit_NOPASS_demo_exec2.R: java.lang.AssertionError: unexpected pending count, expected 1, got 3 at hex.glm.GLM$GLMLambdaTask$LineSearchIteration.callback(GLM.java:529)
glm lbgs needs residual_deviance/residual_degrees_of_freedom/null_deviance/null_degrees_of_freedom in the build_model
GLM model coefficients that are unused (due to lambda or whatever)...model returns ''". Not like h2o1, which didn't return anything.
GBM ModelMetrics, airlines_all (8 machines)*** Attempting to block on task (class water.TaskGetKey) with equal or lower priority. Can lead to deadlock! 122 <= 122
mixed enum plus int: 0 gets parsed as enum, not NA
intermitent GLM2[dest=GLMModel__926238c425adf3e6b8cb7dde15df47de, iteration=9, lambda = 1.0E-5]: invoking line search barrier onExCompletion for hex.glm.GLM$1@2678b3c1 java.lang.AssertionError: unexpected pending count, expected 1, got 2
Runit testing handleSimpleError seems to blow up with something it doesn't like
Add support for standard errors for model predictions
Need to clarify the definition of Typeahead.json for file/folder and limit=parameter
Certificate based authentication for H2O Web for access control.
Integrate and Blogument H2O Scoring Engine POJO into Spark Streaming
GLM build model : For 100 GB file takes forever ...................
GLM lambda search finds fewer non-zeros with decreasing lambda
GLM lambda_max = 0 AssertionError
DRF/GBM can't handle columns with 60% values
DRF can't fit in 320GB cluster memory for 1.33GB Frame
GLM throws AIIOB
Parse in 99% finish state for very long
rapids: h2o1 didn't allow strings in expressions. (enums needed to use encoded value). What does rapids support?
Finish RUnit test for GLM CV
h2o.glm family = "gaussian" ignores link = "log" setting
Long running H2O instances throwing HTTP 500 - Internal Server Error
nans or inf in beta in glm2 (using fuller set of newer params now). poisson, non_negative=1? covtype.20k.data
GLM2 now dynamically recomputes thresholds, but apparently CMs are not in sync..leads to mismatched choice between thresholds and CMs (when this happens the CM displayed won't be right?)
mapr 4.0.1 (which supports yarn and spark) apparently needs a new h2odriver? java.lang.IncompatibleClassChangeError
When a job fails, the model key used in the job remains locked and cannot be reused from R
Double-quoted integers are not parsed as enum
Typo on -ip="..." UnknownHostException() .
In-place feature creation
h2o1 does not encode special html chars before displaying enum. Causes everything after < to be not displayed
Should customers be able to compare wc -l on a dataset, vs the numRows that h2o reports, and match?
Java8 + Eclipse Luna compiling byte code which is rejected by our Weaver
h2o.glm falsely converges when predicted probabilities range broadly between [0, 1] boundaries
Improve NLP capabilities
too many files when you spill keys. Also: spilling seems slow
issue 1 of 222