Remove num_update as param

config/sac_trainer_config.yaml

     max_steps: 5.0e5
     memory_size: 128
     normalize: false
-    num_update: 1
     steps_per_update: 10
     num_layers: 2
     time_horizon: 64

ml-agents/mlagents/trainers/sac/trainer.py

             "init_entcoef",
             "max_steps",
             "normalize",
-            "num_update",
+            "steps_per_update",
             "sequence_length",
             "summary_freq",
             "tau",