Cluster hetrogenious logs having no label using BERT(Unsupervised Learning)

Question

I have a log file that contains different services logs like apache, Hadoop, spark, ssh, HDFS, HPC, and many other types of logs in a single file. I tokenize the logs using BERT