NerDLAproach bestModelMetric "Param bestModelMetric does not exist" Error

[zetobur]

I encountered an error which says "java.util.NoSuchElementException: Param bestModelMetric does not exist." when I try to execute https://nlp.johnsnowlabs.com/docs/en/serving_spark_nlp_via_api_databricks_mlflow example on Databricks.

Then I saw that this metric is newly committed and saying "Make sure bestModelMetric is being used". But at first I couldn't find what the value should be. Then found some explanations (#6749 ) about this and inside them there are values such as "F1 Micro-average, F1 Macro-average, f1_micro, f1_macro, microF1, macrof1".

But whatever I tried all of them one by one inside setBestModelMetric, and setUseBestModel(True) still getting same error.

Here is the code:
`
MODEL_NAME = 'nlp_mlflow'
EXPERIMENT_NAME = 'test_4'
OUTPUT_DIR = f"{MODEL_NAME}_{EXPERIMENT_NAME}_output" # Output folder of all your model artifacts
MODEL_DIR = f"/Users/sgurleye@its.jnj.com/" # Name of the folder where the MLFlow model will be stored

MAX_EPOCHS = 2
LEARNING_RATE = 0.001
BATCH_SIZE = 8
RANDOM_SEED = 0
VALIDATION_SPLIT = 0.2
BESTMODELMETRIC = "microf1"

document = DocumentAssembler()
.setInputCol("text")
.setOutputCol("document")

sentence = SentenceDetector()
.setInputCols(['document'])
.setOutputCol('sentence')

token = Tokenizer()
.setInputCols(['sentence'])
.setOutputCol('token')

embeddings = BertEmbeddings.pretrained("bert_base_cased", "en")
.setInputCols("sentence", "token")
.setOutputCol("embeddings")

ner_approach = NerDLApproach()
.setInputCols(["sentence", "token", "embeddings"])
.setLabelColumn("label")
.setOutputCol("ner")
.setMaxEpochs(MAX_EPOCHS)
.setLr(LEARNING_RATE)
.setBatchSize(BATCH_SIZE)
.setRandomSeed(RANDOM_SEED)
.setVerbose(1)
.setEnableOutputLogs(True)
.setIncludeConfidence(True)
.setIncludeAllConfidenceScores(True)
.setEvaluationLogExtended(True)
.setOutputLogsPath(OUTPUT_DIR)
.setValidationSplit(VALIDATION_SPLIT)
.setUseBestModel(True)
.setBestModelMetric(BESTMODELMETRIC)

ner_preprocessing_pipeline = Pipeline(stages=[
document,
sentence,
token,
embeddings
])

ner_training_pipeline = Pipeline(stages = ner_preprocessing_pipeline.getStages() + [ner_approach])

ner_model = ner_training_pipeline.fit(training_data)
`

Still receiving the same error "java.util.NoSuchElementException: Param bestModelMetric does not exist."

Could you clarify me please? Is there anything wrong in the example? @maziyarpanahi

[maziyarpanahi]

Answered here #7470