setConfigProtoBytes does not work for NerDLModel

[fxu36]

Is there an existing issue for this?

• I have searched the existing issues and did not find a match.

Who can help?

No response

What are you working on?

I have an NER mode in lightweight pipeline and try to configure the TensorFlow parameters through setConfigProtoBytes before inference. The SparkNLP version is 5.2.3

Current Behavior

The TensorFlow parameters set by setConfigProtoBytes are not applied to the TensorFlow session because the load function in NerDLModel already created a session with hard coded configuraion

Expected Behavior

The TensorFlow parameters set by setConfigProtoBytes are not applied to the TensorFlow session because the load function in NerDLModel already created a session with hard coded configuration TensorflowWrapper.TFSessionConfi

Steps To Reproduce

// Load the model
val model = NerDLModel.load()
// The os information
val os = ManagementFactory.getOperatingSystemMXBean()
val cpu = os.getAvailableProcessors()
// ConfigProto
val config = ConfigProto.newBuilder()
.putDeviceCount("GPU", 0)
.putDeviceCount("CPU", cpu)
.setInterOpParallelismThreads(cpu)
.setIntraOpParallelismThreads(cpu)
.setAllowSoftPlacement(true)
.setIsolateSessionState(true)
.setUsePerSessionThreads(true)
.build()
val ner = model.setInputCols("sentence", "token", "embeddings")
.setOutputCol("ner")
.setBatchSize(cpu * 2)
.setConfigProtoBytes(config.toByteArray.map(_.toInt))

Spark NLP version and Apache Spark

SparkNLP 5.2.3
Spache Apark 3.1.2

Type of Spark Application

Scala Application

Java Version

java version "1.8.0_411"

Java Home Directory

C:\Program Files\Java\jdk-1.8

Setup and installation

sbt

Operating System and Version

Windows 2019

Link to your project (if available)

No response

Additional Information

No response

[github-actions]

This issue is stale because it has been open 180 days with no activity. Remove stale label or comment or this will be closed in 5 days