I am trying to load google's Pre-trained vectors 'GoogleNews-vectors-negative300.bin.gz' Google-word2vec into spark.
I converted the bin file to txt and created a smaller chunk for testing that I called 'vectors.txt'. I tried loading it as the following:
val sparkSession = SparkSession.builder
.master("local[*]")
.appName("Word2VecExample")
.getOrCreate()
val model2= Word2VecModel.load(sparkSession.sparkContext, "src/main/resources/vectors.txt")
val synonyms = model2.findSynonyms("the", 5)
for((synonym, cosineSimilarity) <- synonyms) {
println(s"$synonym $cosineSimilarity")
}
and to my surprise I am faced with the following error:
Exception in thread "main" org.apache.hadoop.mapred.InvalidInputException: Input path does not exist: file:/home/elievex/Repository/ARCANA/src/main/resources/vectors.txt/metadata
I'm not sure where did the 'metadata' after 'vectors.txt' came from.
I am using Spark, Scala and Scala IDE for Eclipse.
What am I doing wrong? is there a different way to load a pre-trained model in spark? Would appreciate any tips.
How exactly did you get vector.txt? If you read JavaDoc for Word2VecModel.save you may see that:
This saves: - human-readable (JSON) model metadata to path/metadata/ - Parquet formatted data to path/data/
The model may be loaded using Loader.load.
So what you need is model in Parquet format which is standard for Spark ML models.
Unfortunately load from Google's native format has not been implemented yet (see SPARK-9484).
Related
So, I made an Scala Application to run in Spark, and created the Uber Jar using sbt> assembly.
The file I load is a lookup needed by the application, thus the idea is to package it together. It works fine from within InteliJ using the path "src/main/resources/lookup01.csv"
I am developing in Windows, testing locally, to after deploy it to a remote test server.
But when I call spark-submit on the Windows machine, I get the error :
"org.apache.spark.sql.AnalysisException: Path does not exist: file:/H:/dev/Spark/spark-2.4.3-bin-hadoop2.7/bin/src/main/resources/"
Seems it tries to find the file in the sparkhome location instead of from inside the JAr file.
How could I express the Path so it works looking the file from within the JAR package?
Example code of the way I load the Dataframe. After loading it I transform it into other structures like Maps.
val v_lookup = sparkSession.read.option( "header", true ).csv( "src/main/resources/lookup01.csv")
What I would like to achieve is getting as way to express the path so it works in every environment I try to run the JAR, ideally working also from within InteliJ while developing.
Edit: scala version is 2.11.12
Update:
Seems that to get a hand in the file inside the JAR, I have to read it as a stream, the bellow code worked, but I cant figure out a secure way to extract the headers of the file such as SparkSession.read.option has.
val fileStream = scala.io.Source.getClass.getResourceAsStream("/lookup01.csv")
val inputDF = sparkSession.sparkContext.makeRDD(scala.io.Source.fromInputStream(fileStream).getLines().toList).toDF
When the makeRDD is applied, I get the RDD and then can convert it to a dataframe, but it seems I lost the ability tu use the option from "read" that parsed out the headers as the schema.
Any way around it when using makeRDD ?
Other problem with this is that seems that I will have to manually parse the lines into columns.
You have to get the correct path from classPath
Considering that your file is under src/main/resources:
val path = getClass.getResource("/lookup01.csv")
val v_lookup = sparkSession.read.option( "header", true ).csv(path)
So, it all points to that after the file is inside JAR, it can only be accessed as a inputstream to read the chunk of data from within the compressed file.
I arrived at a solution, even though its not pretty it does what I need, that is to read a csv file, take the 2 first columns and make it into a dataframe and after load it inside a key-value structure (in this case i created a case class to hold these pairs).
I am considering migrating these lookups to a HOCON file, that may make the process less convoluted to load these lookups
import sparkSession.implicits._
val fileStream = scala.io.Source.getClass.getResourceAsStream("/lookup01.csv")
val input = sparkSession.sparkContext.makeRDD(scala.io.Source.fromInputStream(fileStream).getLines().toList).toDF()
val myRdd = input.map {
line =>
val col = utils.Utils.splitCSVString(line.getString(0))
KeyValue(col(0), col(1))
}
val myDF = myRdd.rdd.map(x => (x.key, x.value)).collectAsMap()
fileStream.close()
I have one of our Python systems generating Parquet files using Pandas and fastparquet. These are to be read by a Scala system that runs atop Akka streams.
Akka does provide a source for reading Avro Parquet files. However, when I try to read the file, I end up with
java.lang.IllegalArgumentException: INT96 not yet implemented.
This is one of the columns that does not need to be read for the Scala application to work. My question is whether I can specify a schema and get just that one column out considering that the generated file is from fastparquet.
The relevant snippet which generates a source for reading Parquet files is:
.map(result => {
val path = s"s3a://${result.bucketName}/${result.key}"
val file = HadoopInputFile.fromPath(new Path(path), hadoopConfig)
val reader: ParquetReader[GenericRecord] =
AvroParquetReader
.builder[GenericRecord](file)
.withConf(hadoopConfig)
.build()
AvroParquetSource(reader)
})
In Spark MLlib, BisectingKMeansModel in pyspark have no save/load function.
why?
How to save or load the BisectingKMeans Model with Python to HDFS ?
It may be your spark version. For bisecting k_means is recommended to have above 2.1.0.
You can find a complete example here on the class pyspark.ml.clustering.BisectingKMeans, hope it helps:
https://spark.apache.org/docs/2.1.0/api/python/pyspark.ml.html#pyspark.ml.clustering.BisectingKMeans%20featuresCol=%22features%22,%20predictionCol=%22prediction%22
The last part of the example code include a model save/load:
model_path = temp_path + "/bkm_model"
model.save(model_path)
model2 = BisectingKMeansModel.load(model_path)
It works for hdfs as well, but make sure that temp_path/bkm_model folder does not exist before saving the model or it will give you an error:
(java.io.IOException: Path <temp_path>/bkm_model already exists)
We are using pyspark 1.6. and are trying to convert Text to other file format
(like Json,csv etc) with compression (gzip,lz4,snappy etc). But unable to see compressing working.
Please find the code blow we tried. please help us in pointing the issue in our code else suggest an work around.
Just to add to the question, none of the compressions are working in 1.6, but its working fine in spark 2.X
Option 1:
from pyspark import SparkContext SparkConf
sqlContext.setConf("spark.sql.parquet.compression.codec", "snappy")
df = sqlContext.read.format('parquet').load('hdfs:///user/U1/json_parque_snappy')
df.write.format('json').save('hdfs:///user/U1/parquet_json_snappy')
Option 2:
df = sqlContext.read.format('parquet').load('hdfs:///user/U1/json_parque_snappy')
df.write.format('json').option('codec','com.apache.hadoop.io.compress.SnappyCodec').save('hdfs:///user/U1/parquet_json_snappy_4')
Option 3:
df = sqlContext.read.format('parquet').load('hdfs:///user/U1/json_parque_snappy')
df.write.format('json').option('compression','snappy').save('hdfs:///user/U1/parquet_json_snappy')
For Spark 1.6, to save text/json output, try using the
spark.hadoop.mapred.output.compression.codec parameter
There are 4 parameters to be set. This has been answered already and more details are in this link
With Spark 2.x, the API is simpler and you can use
df.write.option("compression", "gzip")
I am new to Apache Spark. I ran the sample ALS algorithm code present in the examples folder. I gave a csv file as an input. When I use model.save(path) to save the model, it is stored in gz.parquet file.
When I tried to open this file, I get these errors
Now I want to store the recommendation model generated in a text or csv file for using it outside Spark.
I tried the following function to store the model generated in a file but it was useless:
model.saveAsTextFile("path")
Please suggest me a way to overcome this issue.
Lest say you have trained your model with something like this:
val model = ALS.train(ratings, rank, numIterations, 0.01)
All that you have to do is:
import org.apache.spark.mllib.recommendation.ALS
import org.apache.spark.mllib.recommendation.MatrixFactorizationModel
import org.apache.spark.mllib.recommendation.Rating
// Save
model.save(sc, "yourpath/yourmodel")
// Load Model
val sameModel = MatrixFactorizationModel.load(sc, "yourpath/yourmodel")
As it turns out saveAsTextFile() only works on the slaves.Use collect() to collect the data from the slaves so it can be saved locally on the master. Solution can be found here