AccessDenied errors loading healthcare models

When creating annotators from pretrained models in versions 2.7.4 and 2.7.5 we are getting AccessDenied errors when running Spark locally (master = “local”). Version 2.7.3 works fine and running 2.7.4/2.7.5 on AWS EMR cluster also works.

We are specifying the AWS credentials in the .aws/credentials file using [spark_nlp] profile.

Current Behavior

We get the following error when trying to load a pretrained model from the clinical/models location.

Py4JJavaError: An error occurred while calling z:com.johnsnowlabs.nlp.pretrained.PythonResourceDownloader.getDownloadSize.
: com.amazonaws.services.s3.model.AmazonS3Exception: Status Code: 403, AWS Service: Amazon S3, AWS Request ID: 7JGS2EJM96C6SHGG, AWS Error Code: AccessDenied, AWS Error Message: Access Denied, S3 Extended Request ID: KgsQ87K74+KaiKhc7arCbcuVtc8POle+iUPrzaJ2XG5ECu37HiN5VDNzqKBs4DunvOG1yqEJlwg=
	at com.amazonaws.http.AmazonHttpClient.handleErrorResponse(AmazonHttpClient.java:798)
	at com.amazonaws.http.AmazonHttpClient.executeHelper(AmazonHttpClient.java:421)
	at com.amazonaws.http.AmazonHttpClient.execute(AmazonHttpClient.java:232)
	at com.amazonaws.services.s3.AmazonS3Client.invoke(AmazonS3Client.java:3528)
	at com.amazonaws.services.s3.AmazonS3Client.getObject(AmazonS3Client.java:1111)
	at com.amazonaws.services.s3.AmazonS3Client.getObject(AmazonS3Client.java:984)
	at com.johnsnowlabs.nlp.pretrained.S3ResourceDownloader.downloadMetadataIfNeed(S3ResourceDownloader.scala:69)
	at com.johnsnowlabs.nlp.pretrained.S3ResourceDownloader.resolveLink(S3ResourceDownloader.scala:81)
	at com.johnsnowlabs.nlp.pretrained.S3ResourceDownloader.getDownloadSize(S3ResourceDownloader.scala:159)
	at com.johnsnowlabs.nlp.pretrained.ResourceDownloader$.getDownloadSize(ResourceDownloader.scala:403)
	at com.johnsnowlabs.nlp.pretrained.PythonResourceDownloader$.getDownloadSize(ResourceDownloader.scala:501)
	at com.johnsnowlabs.nlp.pretrained.PythonResourceDownloader.getDownloadSize(ResourceDownloader.scala)
	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 py4j.reflection.MethodInvoker.invoke(MethodInvoker.java:244)
	at py4j.reflection.ReflectionEngine.invoke(ReflectionEngine.java:357)
	at py4j.Gateway.invoke(Gateway.java:282)
	at py4j.commands.AbstractCommand.invokeMethod(AbstractCommand.java:132)
	at py4j.commands.CallCommand.execute(CallCommand.java:79)
	at py4j.GatewayConnection.run(GatewayConnection.java:238)
	at java.lang.Thread.run(Thread.java:748)

Steps to Reproduce

import sparknlp
from pyspark.sql import SparkSession

spark = SparkSession.builder \
    .appName("Spark NLP")\
    .master("local[4]")\
    .config("spark.driver.memory","16G")\
    .config("spark.driver.maxResultSize", "0") \
    .config("spark.jars.packages", "com.johnsnowlabs.nlp:spark-nlp_2.11:2.7.5")\
    .config("spark.kryoserializer.buffer.max", "1000M")\
    .getOrCreate()

from sparknlp.annotator import *

sent_detect = SentenceDetectorDLModel() \
  .pretrained("sentence_detector_dl_healthcare", "en", "clinical/models") \
  .setInputCols(["document"]) \
  .setOutputCol("sentences")

Context

Your Environment

• Spark NLP version sparknlp.version(): 2.7.4
• Apache NLP version spark.version: 2.4.5
• Java version java -version: openjdk version “1.8.0_282”

[dkincaid]

I tried the new 3.0.0 version that was released today and this issue seems to be fixed in that version.

[dkincaid]

Version 2.7.3 and earlier of the Spark NLP jar works with AWS credentials stored in the .aws/credentials file. The problem with specifying them in environment variables is that it only allows one set of credentials to be in effect at a time. We can’t do that since we need our own credentials to access our AWS resources. As far as I know the only way to have different credentials setup is to use profiles which is what the .aws/credentials file provides. I suspect the problem with 2.7.4 and 2.7.5 has to do with reverting to a really old version of the AWS SDK (1.7.4). Version 2.7.3 uses 1.11.603. You were using 1.11.603 at least as far back as 2.5.0.