logging

[gromajus]

Hi!
I would like to ask if it is possible to turn off logging or change the logging level from python script that uses nlu library?
Even simple 'import nlu' generates lines of logs, loading models there are tons of them...

Before importing nlu, I am trying to create pyspark context and set desired log level as it is pointed in logs from import nlu: Setting default log level to "WARN".
To adjust logging level use sc.setLogLevel(newLevel)

But it doesn't seem to help, actually the opposite, I can't load models and do the predictions then...

The other approach was setting logging levels for all possible loggers: nlu, py4j, py4j.java_gateway to CRITICAL in my case

logging.getLogger('nlu').setLevel(logging.CRITICAL)
logging.getLogger('py4j').setLevel(logging.CRITICAL)
logging.getLogger('py4j.java_gateway').setLevel(logging.CRITICAL)

But it also didn't help.
There are still messages from e.g. WARN SparkSession$Builder, WARN ApacheUtils, I tensorflow/core/platform/cpu_feature_guard.cc:142], etc...

[C-K-Loan]

HI @gromajus currently there is no straightforward way to mute the Spark Logging messages.

We are looking into this and hopefully can provide a solution soon.

[gromajus]

OK, thank you @C-K-Loan :)