我在per_feature=TRUE的情况下运行H2O异常，这会导致Java堆空间错误。在其他一些关于这个错误消息的帖子中，我看到有人建议使用h2o.remove(df(来释放使用过的内存。然而，在我的情况下，我没有任何循环，而且似乎没有什么可以删除的东西来释放一些用过的内存。
这是我的代码：
library(h2o)
h2o.init(min_mem_size = "10G", max_mem_size = "15G")
data.hex <- as.h2o(data)
x <- names(data.hex)
random_seed <- 42
# Deeplearning Model
print("Deep learning model begins ...")
model.dl = h2o.deeplearning(x = x,
training_frame = data.hex,
autoencoder = TRUE,
activation = "Tanh",
hidden = c(5, 5, 5, 5, 5),
mini_batch_size = 64,
epochs = 100,
stopping_rounds = 15,
variable_importances = TRUE,
seed = random_seed)
# Calculating anomaly per feature
print('Calculating anomaly per feature ...')
errors_per_feature <- h2o.anomaly(model.dl, data.hex, per_feature = TRUE) # Anomaly Detection Algorithm
print('Converting from H2O frame to dataframe ...')
errors1_per_feature <- as.data.frame(errors_per_feature) # Convert back to data frame
以下是详细的错误消息：
[1] "Deep learning model begins ..."
|======================================================================| 100%
[1] "Calculating anomaly per feature ..."
ERROR: Unexpected HTTP Status code: 500 Server Error (url = http://localhost:54321/3/Predictions/models/DeepLearning_model_R_1594826474037_2/frames/Accesses_sid_a71f_1)
water.util.DistributedException
[1] "DistributedException from localhost/127.0.0.1:54321: 'Java heap space', caused by java.lang.OutOfMemoryError: Java heap space"
[2] " water.MRTask.getResult(MRTask.java:494)"
[3] " water.MRTask.getResult(MRTask.java:502)"
[4] " water.MRTask.doAll(MRTask.java:397)"
[5] " water.MRTask.doAll(MRTask.java:403)"
[6] " hex.deeplearning.DeepLearningModel.scoreAutoEncoder(DeepLearningModel.java:761)"
[7] " water.api.ModelMetricsHandler.predict(ModelMetricsHandler.java:469)"
[8] " java.base/jdk.internal.reflect.NativeMethodAccessorImpl.invoke0(Native Method)"
[9] " java.base/jdk.internal.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)"
[10] " java.base/jdk.internal.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)"
[11] " java.base/java.lang.reflect.Method.invoke(Method.java:567)"
[12] " water.api.Handler.handle(Handler.java:60)"
[13] " water.api.RequestServer.serve(RequestServer.java:470)"
[14] " water.api.RequestServer.doGeneric(RequestServer.java:301)"
[15] " water.api.RequestServer.doPost(RequestServer.java:227)"
[16] " javax.servlet.http.HttpServlet.service(HttpServlet.java:755)"
[17] " javax.servlet.http.HttpServlet.service(HttpServlet.java:848)"
[18] " org.eclipse.jetty.servlet.ServletHolder.handle(ServletHolder.java:684)"
[19] " org.eclipse.jetty.servlet.ServletHandler.doHandle(ServletHandler.java:501)"
[20] " org.eclipse.jetty.server.handler.ContextHandler.doHandle(ContextHandler.java:1086)"
[21] " org.eclipse.jetty.servlet.ServletHandler.doScope(ServletHandler.java:427)"
[22] " org.eclipse.jetty.server.handler.ContextHandler.doScope(ContextHandler.java:1020)"
[23] " org.eclipse.jetty.server.handler.ScopedHandler.handle(ScopedHandler.java:135)"
[24] " org.eclipse.jetty.server.handler.HandlerCollection.handle(HandlerCollection.java:154)"
[25] " org.eclipse.jetty.server.handler.HandlerWrapper.handle(HandlerWrapper.java:116)"
[26] " water.webserver.jetty8.Jetty8ServerAdapter$LoginHandler.handle(Jetty8ServerAdapter.java:119)"
[27] " org.eclipse.jetty.server.handler.HandlerCollection.handle(HandlerCollection.java:154)"
[28] " org.eclipse.jetty.server.handler.HandlerWrapper.handle(HandlerWrapper.java:116)"
[29] " org.eclipse.jetty.server.Server.handle(Server.java:370)"
[30] " org.eclipse.jetty.server.AbstractHttpConnection.handleRequest(AbstractHttpConnection.java:494)"
[31] " org.eclipse.jetty.server.BlockingHttpConnection.handleRequest(BlockingHttpConnection.java:53)"
[32] " org.eclipse.jetty.server.AbstractHttpConnection.content(AbstractHttpConnection.java:984)"
[33] " org.eclipse.jetty.server.AbstractHttpConnection$RequestHandler.content(AbstractHttpConnection.java:1045)"
[34] " org.eclipse.jetty.http.HttpParser.parseNext(HttpParser.java:861)"
[35] " org.eclipse.jetty.http.HttpParser.parseAvailable(HttpParser.java:236)"
[36] " org.eclipse.jetty.server.BlockingHttpConnection.handle(BlockingHttpConnection.java:72)"
[37] " org.eclipse.jetty.server.bio.SocketConnector$ConnectorEndPoint.run(SocketConnector.java:264)"
[38] " org.eclipse.jetty.util.thread.QueuedThreadPool.runJob(QueuedThreadPool.java:608)"
[39] " org.eclipse.jetty.util.thread.QueuedThreadPool$3.run(QueuedThreadPool.java:543)"
[40] " java.base/java.lang.Thread.run(Thread.java:830)"
[41] "Caused by:java.lang.OutOfMemoryError: Java heap space"
Error in .h2o.doSafeREST(h2oRestApiVersion = h2oRestApiVersion, urlSuffix = page, :
ERROR MESSAGE:
DistributedException from localhost/127.0.0.1:54321: 'Java heap space'
Calls: h2o.anomaly -> .h2o.__remoteSend -> .h2o.doSafeREST
Execution halted
R和H2O版本：
H2O cluster version: 3.30.0.6
H2O cluster total nodes: 1
H2O cluster total memory: 13.43 GB
H2O cluster total cores: 16
H2O cluster allowed cores: 16
H2O cluster healthy: TRUE
R Version: R version 3.6.3 (2020-02-29)
我的macOS上有16GB的内存。
数据中有6个变量(列(：5个分类变量和1个数字变量。5个分类变量中的唯一值数量分别为17、49、52、85和5032。行数约为500k。数据文件大小为44 MB(在H2O中编码之前(。
在我的情况下，我能做些什么来解决这个问题？如果我能提供任何其他信息，请告诉我。谢谢你的帮助！
[也在这里剪切并粘贴我对h2o流邮件列表的响应…]
我怀疑大量的分类层次导致记忆爆炸。
试着删除那个变量，看看它是否至少完成了。
如果是这样的话，试着以某种方式将其重新合并到较少的级别中。