def test_autologging_dedups_multiple_reads_of_same_datasource(
spark_session, format_to_file_path):
mlflow.spark.autolog()
data_format = list(format_to_file_path.keys())[0]
file_path = format_to_file_path[data_format]
df = (spark_session.read.format(data_format).option(
"header", "true").option("inferSchema", "true").load(file_path))
with mlflow.start_run():
run_id = mlflow.active_run().info.run_id
df.collect()
df.filter("number1 > 0").collect()
df.limit(2).collect()
df.collect()
time.sleep(1)
run = mlflow.get_run(run_id)
_assert_spark_data_logged(run=run, path=file_path, data_format=data_format)
# Test context provider flow
df.filter("number1 > 0").collect()
df.limit(2).collect()
df.collect()
with mlflow.start_run():
run_id2 = mlflow.active_run().info.run_id
time.sleep(1)
run2 = mlflow.get_run(run_id2)
_assert_spark_data_logged(run=run2,
path=file_path,
data_format=data_format)
def test_autologging_of_datasources_with_different_formats(
spark_session, format_to_file_path):
mlflow.spark.autolog()
for data_format, file_path in format_to_file_path.items():
base_df = (spark_session.read.format(data_format).option(
"header", "true").option("inferSchema", "true").load(file_path))
base_df.createOrReplaceTempView("temptable")
table_df0 = spark_session.table("temptable")
table_df1 = spark_session.sql(
"SELECT number1, number2 from temptable LIMIT 5")
dfs = [
base_df,
table_df0,
table_df1,
base_df.filter("number1 > 0"),
base_df.select("number1"),
base_df.limit(2),
base_df.filter("number1 > 0").select("number1").limit(2),
]
for df in dfs:
with mlflow.start_run():
run_id = mlflow.active_run().info.run_id
df.collect()
time.sleep(1)
run = mlflow.get_run(run_id)
_assert_spark_data_logged(run=run,
path=file_path,
data_format=data_format)
def test_enabling_autologging_before_spark_session_works(disable):
mlflow.spark.autolog(disable=disable)
# creating spark session AFTER autolog was enabled
spark_session = _get_or_create_spark_session()
rows = [Row(100)]
schema = StructType([StructField("number2", IntegerType())])
rdd = spark_session.sparkContext.parallelize(rows)
df = spark_session.createDataFrame(rdd, schema)
tempdir = tempfile.mkdtemp()
filepath = os.path.join(tempdir, "test-data")
df.write.option("header", "true").format("csv").save(filepath)
read_df = (spark_session.read.format("csv").option(
"header", "true").option("inferSchema", "true").load(filepath))
with mlflow.start_run():
run_id = mlflow.active_run().info.run_id
read_df.collect()
time.sleep(1)
run = mlflow.get_run(run_id)
if disable:
_assert_spark_data_not_logged(run=run)
else:
_assert_spark_data_logged(run=run, path=filepath, data_format="csv")
shutil.rmtree(tempdir)
spark_session.stop()
def test_spark_autologging_with_sklearn_autologging(spark_session, data_format,
file_path):
assert mlflow.active_run() is None
mlflow.spark.autolog()
mlflow.sklearn.autolog()
df = (spark_session.read.format(data_format).option(
"header",
"true").option("inferSchema",
"true").load(file_path).select("number1", "number2"))
pandas_df = df.toPandas()
run = _fit_sklearn_model(pandas_df)
_assert_spark_data_logged(run, file_path, data_format)
assert mlflow.active_run() is None
def test_autologging_multiple_runs_same_data(spark_session,
format_to_file_path):
mlflow.spark.autolog()
data_format = list(format_to_file_path.keys())[0]
file_path = format_to_file_path[data_format]
df = (spark_session.read.format(data_format).option(
"header", "true").option("inferSchema", "true").load(file_path))
df.collect()
for _ in range(2):
with mlflow.start_run():
time.sleep(1)
run_id = mlflow.active_run().info.run_id
run = mlflow.get_run(run_id)
_assert_spark_data_logged(run=run,
path=file_path,
data_format=data_format)
def test_autologging_disabled_then_enabled(spark_session, format_to_file_path):
mlflow.spark.autolog(disable=True)
data_format = list(format_to_file_path.keys())[0]
file_path = format_to_file_path[data_format]
df = (spark_session.read.format(data_format).option(
"header", "true").option("inferSchema", "true").load(file_path))
# Logging is disabled here.
with mlflow.start_run():
run_id = mlflow.active_run().info.run_id
df.collect()
time.sleep(1)
run = mlflow.get_run(run_id)
_assert_spark_data_not_logged(run=run)
# Logging is enabled here.
mlflow.spark.autolog(disable=False)
with mlflow.start_run():
run_id = mlflow.active_run().info.run_id
df.filter("number1 > 0").collect()
time.sleep(1)
run = mlflow.get_run(run_id)
_assert_spark_data_logged(run=run, path=file_path, data_format=data_format)
def test_spark_sklearn_autologging_context_provider(spark_session, data_format,
file_path):
mlflow.spark.autolog()
mlflow.sklearn.autolog()
df = (spark_session.read.format(data_format).option(
"header",
"true").option("inferSchema",
"true").load(file_path).select("number1", "number2"))
pandas_df = df.toPandas()
# DF info should be logged to the first run (it should be added to our context provider after
# the toPandas() call above & then logged here)
with mlflow.start_run():
run = _fit_sklearn_model(pandas_df)
_assert_spark_data_logged(run, file_path, data_format)
with mlflow.start_run():
pandas_df2 = df.filter("number1 > 0").toPandas()
run2 = _fit_sklearn_model(pandas_df2)
assert run2.info.run_id != run.info.run_id
_assert_spark_data_logged(run2, file_path, data_format)
time.sleep(1)
assert mlflow.active_run() is None