pyflink矢量化udf抛出nullpointerexception

klsxnrf1  于 2021-07-15  发布在  Flink
关注(0)|答案(1)|浏览(499)

我有一个ml模型需要两个numpy.ndarray- users 以及 items -并返回numpy.ndarray predictions . 在普通的python代码中，我会：

model = load_model()

df = load_data() # the DataFrame includes 4 columns, namely, user_id, movie_id, rating, and timestamp

users = df.user_id.values
items = df.movie_id.values

predictions = model(users, items)

我正在研究将此代码移植到flink中，以利用其分布式特性。我的假设是：通过在多个flink节点上分配预测工作负载，我应该能够更快地运行整个预测。
所以我写了一份皮Flink的工作。注意我实现了一个名为 predict 进行预测。


# batch_prediction.py

model = load_model()

settings = EnvironmentSettings.new_instance().use_blink_planner().build()
exec_env = StreamExecutionEnvironment.get_execution_environment()
t_env = StreamTableEnvironment.create(exec_env, environment_settings=settings)

SOURCE_DDL = """
CREATE TABLE source (
    user_id INT,
    movie_id INT,
    rating TINYINT,
    event_ms BIGINT
) WITH (
    'connector' = 'filesystem',
    'format' = 'csv',
    'csv.field-delimiter' = '\t',
    'path' = 'ml-100k/u1.test'
)
"""

SINK_DDL = """
CREATE TABLE sink (
    prediction DOUBLE
) WITH (
    'connector' = 'print'
)
"""

t_env.execute_sql(SOURCE_DDL)
t_env.execute_sql(SINK_DDL)
t_env.execute_sql(
    "INSERT INTO sink SELECT PREDICT(user_id, movie_id) FROM source"
).wait()

这是自定义项。


# batch_prediction.py (cont)

@udf(result_type=DataTypes.DOUBLE())
def predict(user, item):
    return model([user], [item]).item()

t_env.create_temporary_function("predict", predict)

工作进展顺利。然而，这个预测实际上运行在数据的每一行上 source 表，未执行。相反，我想把80000（user\u id，movie\u id）对分成100个批，每个批有800行。作业触发 model(users, items) 函数100次（=#批），其中 users 以及 items 有800个元素。
我找不到办法。通过查看文档，矢量化的用户定义函数可能会起作用。


# batch_prediction.py (snippet)

# I add the func_type="pandas"

@udf(result_type=DataTypes.DOUBLE(), func_type="pandas")
def predict(user, item):
    ...

不幸的是，事实并非如此。

> python batch_prediction.py
...
Traceback (most recent call last):
  File "batch_prediction.py", line 55, in <module>
    "INSERT INTO sink SELECT PREDICT(user_id, movie_id) FROM source"
  File "/usr/local/anaconda3/envs/flink-ml/lib/python3.7/site-packages/pyflink/table/table_result.py", line 76, in wait
    get_method(self._j_table_result, "await")()
  File "/usr/local/anaconda3/envs/flink-ml/lib/python3.7/site-packages/py4j/java_gateway.py", line 1286, in __call__
    answer, self.gateway_client, self.target_id, self.name)
  File "/usr/local/anaconda3/envs/flink-ml/lib/python3.7/site-packages/pyflink/util/exceptions.py", line 147, in deco
    return f(*a,**kw)
  File "/usr/local/anaconda3/envs/flink-ml/lib/python3.7/site-packages/py4j/protocol.py", line 328, in get_return_value
    format(target_id, ".", name), value)
py4j.protocol.Py4JJavaError: An error occurred while calling o51.await.
: java.util.concurrent.ExecutionException: org.apache.flink.table.api.TableException: Failed to wait job finish
    at java.util.concurrent.CompletableFuture.reportGet(CompletableFuture.java:357)
    at java.util.concurrent.CompletableFuture.get(CompletableFuture.java:1908)
    at org.apache.flink.table.api.internal.TableResultImpl.awaitInternal(TableResultImpl.java:119)
    at org.apache.flink.table.api.internal.TableResultImpl.await(TableResultImpl.java:86)
    at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
    at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
    at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
    at java.lang.reflect.Method.invoke(Method.java:498)
    at org.apache.flink.api.python.shaded.py4j.reflection.MethodInvoker.invoke(MethodInvoker.java:244)
    at org.apache.flink.api.python.shaded.py4j.reflection.ReflectionEngine.invoke(ReflectionEngine.java:357)
    at org.apache.flink.api.python.shaded.py4j.Gateway.invoke(Gateway.java:282)
    at org.apache.flink.api.python.shaded.py4j.commands.AbstractCommand.invokeMethod(AbstractCommand.java:132)
    at org.apache.flink.api.python.shaded.py4j.commands.CallCommand.execute(CallCommand.java:79)
    at org.apache.flink.api.python.shaded.py4j.GatewayConnection.run(GatewayConnection.java:238)
    at java.lang.Thread.run(Thread.java:748)
Caused by: org.apache.flink.table.api.TableException: Failed to wait job finish
    at org.apache.flink.table.api.internal.InsertResultIterator.hasNext(InsertResultIterator.java:59)
    at org.apache.flink.table.api.internal.TableResultImpl$CloseableRowIteratorWrapper.hasNext(TableResultImpl.java:355)
    at org.apache.flink.table.api.internal.TableResultImpl$CloseableRowIteratorWrapper.isFirstRowReady(TableResultImpl.java:368)
    at org.apache.flink.table.api.internal.TableResultImpl.lambda$awaitInternal$1(TableResultImpl.java:107)
    at java.util.concurrent.CompletableFuture$AsyncRun.run(CompletableFuture.java:1640)
    at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
    at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
    ... 1 more
Caused by: java.util.concurrent.ExecutionException: org.apache.flink.runtime.client.JobExecutionException: Job execution failed.
    at java.util.concurrent.CompletableFuture.reportGet(CompletableFuture.java:357)
    at java.util.concurrent.CompletableFuture.get(CompletableFuture.java:1908)
    at org.apache.flink.table.api.internal.InsertResultIterator.hasNext(InsertResultIterator.java:57)
    ... 7 more
Caused by: org.apache.flink.runtime.client.JobExecutionException: Job execution failed.
    at org.apache.flink.runtime.jobmaster.JobResult.toJobExecutionResult(JobResult.java:147)
    at org.apache.flink.runtime.minicluster.MiniClusterJobClient.lambda$getJobExecutionResult$2(MiniClusterJobClient.java:119)
    at java.util.concurrent.CompletableFuture.uniApply(CompletableFuture.java:616)
    at java.util.concurrent.CompletableFuture$UniApply.tryFire(CompletableFuture.java:591)
    at java.util.concurrent.CompletableFuture.postComplete(CompletableFuture.java:488)
    at java.util.concurrent.CompletableFuture.complete(CompletableFuture.java:1975)
    at org.apache.flink.runtime.rpc.akka.AkkaInvocationHandler.lambda$invokeRpc$0(AkkaInvocationHandler.java:229)
    at java.util.concurrent.CompletableFuture.uniWhenComplete(CompletableFuture.java:774)
    at java.util.concurrent.CompletableFuture$UniWhenComplete.tryFire(CompletableFuture.java:750)
    at java.util.concurrent.CompletableFuture.postComplete(CompletableFuture.java:488)
    at java.util.concurrent.CompletableFuture.complete(CompletableFuture.java:1975)
    at org.apache.flink.runtime.concurrent.FutureUtils$1.onComplete(FutureUtils.java:996)
    at akka.dispatch.OnComplete.internal(Future.scala:264)
    at akka.dispatch.OnComplete.internal(Future.scala:261)
    at akka.dispatch.japi$CallbackBridge.apply(Future.scala:191)
    at akka.dispatch.japi$CallbackBridge.apply(Future.scala:188)
    at scala.concurrent.impl.CallbackRunnable.run(Promise.scala:36)
    at org.apache.flink.runtime.concurrent.Executors$DirectExecutionContext.execute(Executors.java:74)
    at scala.concurrent.impl.CallbackRunnable.executeWithValue(Promise.scala:44)
    at scala.concurrent.impl.Promise$DefaultPromise.tryComplete(Promise.scala:252)
    at akka.pattern.PromiseActorRef.$bang(AskSupport.scala:572)
    at akka.pattern.PipeToSupport$PipeableFuture$$anonfun$pipeTo$1.applyOrElse(PipeToSupport.scala:22)
    at akka.pattern.PipeToSupport$PipeableFuture$$anonfun$pipeTo$1.applyOrElse(PipeToSupport.scala:21)
    at scala.concurrent.Future$$anonfun$andThen$1.apply(Future.scala:436)
    at scala.concurrent.Future$$anonfun$andThen$1.apply(Future.scala:435)
    at scala.concurrent.impl.CallbackRunnable.run(Promise.scala:36)
    at akka.dispatch.BatchingExecutor$AbstractBatch.processBatch(BatchingExecutor.scala:55)
    at akka.dispatch.BatchingExecutor$BlockableBatch$$anonfun$run$1.apply$mcV$sp(BatchingExecutor.scala:91)
    at akka.dispatch.BatchingExecutor$BlockableBatch$$anonfun$run$1.apply(BatchingExecutor.scala:91)
    at akka.dispatch.BatchingExecutor$BlockableBatch$$anonfun$run$1.apply(BatchingExecutor.scala:91)
    at scala.concurrent.BlockContext$.withBlockContext(BlockContext.scala:72)
    at akka.dispatch.BatchingExecutor$BlockableBatch.run(BatchingExecutor.scala:90)
    at akka.dispatch.TaskInvocation.run(AbstractDispatcher.scala:40)
    at akka.dispatch.ForkJoinExecutorConfigurator$AkkaForkJoinTask.exec(ForkJoinExecutorConfigurator.scala:44)
    at akka.dispatch.forkjoin.ForkJoinTask.doExec(ForkJoinTask.java:260)
    at akka.dispatch.forkjoin.ForkJoinPool$WorkQueue.runTask(ForkJoinPool.java:1339)
    at akka.dispatch.forkjoin.ForkJoinPool.runWorker(ForkJoinPool.java:1979)
    at akka.dispatch.forkjoin.ForkJoinWorkerThread.run(ForkJoinWorkerThread.java:107)
Caused by: org.apache.flink.runtime.JobException: Recovery is suppressed by NoRestartBackoffTimeStrategy
    at org.apache.flink.runtime.executiongraph.failover.flip1.ExecutionFailureHandler.handleFailure(ExecutionFailureHandler.java:116)
    at org.apache.flink.runtime.executiongraph.failover.flip1.ExecutionFailureHandler.getFailureHandlingResult(ExecutionFailureHandler.java:78)
    at org.apache.flink.runtime.scheduler.DefaultScheduler.handleTaskFailure(DefaultScheduler.java:224)
    at org.apache.flink.runtime.scheduler.DefaultScheduler.maybeHandleTaskFailure(DefaultScheduler.java:217)
    at org.apache.flink.runtime.scheduler.DefaultScheduler.updateTaskExecutionStateInternal(DefaultScheduler.java:208)
    at org.apache.flink.runtime.scheduler.SchedulerBase.updateTaskExecutionState(SchedulerBase.java:610)
    at org.apache.flink.runtime.scheduler.SchedulerNG.updateTaskExecutionState(SchedulerNG.java:89)
    at org.apache.flink.runtime.jobmaster.JobMaster.updateTaskExecutionState(JobMaster.java:419)
    at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
    at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
    at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
    at java.lang.reflect.Method.invoke(Method.java:498)
    at org.apache.flink.runtime.rpc.akka.AkkaRpcActor.handleRpcInvocation(AkkaRpcActor.java:286)
    at org.apache.flink.runtime.rpc.akka.AkkaRpcActor.handleRpcMessage(AkkaRpcActor.java:201)
    at org.apache.flink.runtime.rpc.akka.FencedAkkaRpcActor.handleRpcMessage(FencedAkkaRpcActor.java:74)
    at org.apache.flink.runtime.rpc.akka.AkkaRpcActor.handleMessage(AkkaRpcActor.java:154)
    at akka.japi.pf.UnitCaseStatement.apply(CaseStatements.scala:26)
    at akka.japi.pf.UnitCaseStatement.apply(CaseStatements.scala:21)
    at scala.PartialFunction$class.applyOrElse(PartialFunction.scala:123)
    at akka.japi.pf.UnitCaseStatement.applyOrElse(CaseStatements.scala:21)
    at scala.PartialFunction$OrElse.applyOrElse(PartialFunction.scala:170)
    at scala.PartialFunction$OrElse.applyOrElse(PartialFunction.scala:171)
    at scala.PartialFunction$OrElse.applyOrElse(PartialFunction.scala:171)
    at akka.actor.Actor$class.aroundReceive(Actor.scala:517)
    at akka.actor.AbstractActor.aroundReceive(AbstractActor.scala:225)
    at akka.actor.ActorCell.receiveMessage(ActorCell.scala:592)
    at akka.actor.ActorCell.invoke(ActorCell.scala:561)
    at akka.dispatch.Mailbox.processMailbox(Mailbox.scala:258)
    at akka.dispatch.Mailbox.run(Mailbox.scala:225)
    at akka.dispatch.Mailbox.exec(Mailbox.scala:235)
    ... 4 more
Caused by: org.apache.flink.streaming.runtime.tasks.AsynchronousException: Caught exception while processing timer.
    at org.apache.flink.streaming.runtime.tasks.StreamTask$StreamTaskAsyncExceptionHandler.handleAsyncException(StreamTask.java:1108)
    at org.apache.flink.streaming.runtime.tasks.StreamTask.handleAsyncException(StreamTask.java:1082)
    at org.apache.flink.streaming.runtime.tasks.StreamTask.invokeProcessingTimeCallback(StreamTask.java:1213)
    at org.apache.flink.streaming.runtime.tasks.StreamTask.lambda$null$17(StreamTask.java:1202)
    at org.apache.flink.streaming.runtime.tasks.StreamTaskActionExecutor$SynchronizedStreamTaskActionExecutor.runThrowing(StreamTaskActionExecutor.java:92)
    at org.apache.flink.streaming.runtime.tasks.mailbox.Mail.run(Mail.java:78)
    at org.apache.flink.streaming.runtime.tasks.mailbox.MailboxExecutorImpl.tryYield(MailboxExecutorImpl.java:91)
    at org.apache.flink.streaming.runtime.tasks.StreamOperatorWrapper.quiesceTimeServiceAndCloseOperator(StreamOperatorWrapper.java:155)
    at org.apache.flink.streaming.runtime.tasks.StreamOperatorWrapper.close(StreamOperatorWrapper.java:130)
    at org.apache.flink.streaming.runtime.tasks.OperatorChain.closeOperators(OperatorChain.java:412)
    at org.apache.flink.streaming.runtime.tasks.StreamTask.afterInvoke(StreamTask.java:585)
    at org.apache.flink.streaming.runtime.tasks.StreamTask.invoke(StreamTask.java:547)
    at org.apache.flink.runtime.taskmanager.Task.doRun(Task.java:722)
    at org.apache.flink.runtime.taskmanager.Task.run(Task.java:547)
    at java.lang.Thread.run(Thread.java:748)
Caused by: TimerException{java.lang.RuntimeException: Failed to close remote bundle}
    ... 13 more
Caused by: java.lang.RuntimeException: Failed to close remote bundle
    at org.apache.flink.streaming.api.runners.python.beam.BeamPythonFunctionRunner.finishBundle(BeamPythonFunctionRunner.java:371)
    at org.apache.flink.streaming.api.runners.python.beam.BeamPythonFunctionRunner.flush(BeamPythonFunctionRunner.java:325)
    at org.apache.flink.streaming.api.operators.python.AbstractPythonFunctionOperator.invokeFinishBundle(AbstractPythonFunctionOperator.java:291)
    at org.apache.flink.table.runtime.operators.python.scalar.arrow.RowDataArrowPythonScalarFunctionOperator.invokeFinishBundle(RowDataArrowPythonScalarFunctionOperator.java:77)
    at org.apache.flink.streaming.api.operators.python.AbstractPythonFunctionOperator.checkInvokeFinishBundleByTime(AbstractPythonFunctionOperator.java:285)
    at org.apache.flink.streaming.api.operators.python.AbstractPythonFunctionOperator.lambda$open$0(AbstractPythonFunctionOperator.java:134)
    at org.apache.flink.streaming.runtime.tasks.StreamTask.invokeProcessingTimeCallback(StreamTask.java:1211)
    ... 12 more
Caused by: java.lang.NullPointerException
    at org.apache.flink.streaming.api.runners.python.beam.BeamPythonFunctionRunner.finishBundle(BeamPythonFunctionRunner.java:369)
    ... 18 more

错误消息不是很有用。有人能帮忙吗？谢谢！
注意：源代码可以在这里找到。要运行代码，您将需要本地anaconda，然后：

conda env create -f environment.yml
conda activate flink-ml
apache-flinkflink-sqlpyflink

1条答案

按热度按时间
atmip9wb

atmip9wb1#

Apache·Flink社区的功劳。参见线程。
对于pandas udf，每个输入参数的输入类型都是pandas.series，结果类型也应该是pandas.series。此外，结果的长度应与输入的长度相同。您能检查一下您的Pandas自定义项实施是否是这样吗？
然后我决定添加一个 pytest 对我的udf进行单元测试，以验证输入和输出类型。以下是方法：

import pandas as pd

from udf_def import predict

def test_predict():
    f = predict._func
    users = pd.Series([1, 2, 3])
    items = pd.Series([1, 4, 9])
    preds = f(users, items)
    assert isinstance(preds, pd.Series)
    assert len(preds) == 3

它揭示了实现错误。然后我修复我的udf实现以通过单元测试。

from model_def import MatrixFactorization

@udf(result_type=DataTypes.DOUBLE(), func_type="pandas")
def predict(users, items):
    n_users, n_items = 943, 1682
    model = MatrixFactorization(n_users, n_items)
    model.load_state_dict(torch.load("model.pth"))
    return pd.Series(model(users, items).detach().numpy())

使用新的udf实现，我可以运行批处理预测。

> python batch_prediction.py
1> +I(-2.3802385330200195)
1> +I(20.000154495239258)
1> +I(-30.704544067382812)
1> +I(-11.602800369262695)
1> +I(10.998968124389648)
...

更新的源代码可以在这里找到。

赞(0）分享  回复(0）举报  2021-07-15
我来回答

相关问题

    查看更多

    热门标签

    更多
    JavaquerypythonNode开发语言requestUtil数据库Table后端算法LoggerMessageElementParser

    最新问答

    更多
    • 蜀ICP备2022004421号-2 NULL123 https://www.null123.com © All rights reserved
    • 本站内容来源互联网,如果侵犯您的权益请联系我们删除, 联系方式：448109455@qq.com