This is a fine-tune of Goliath-120B.
Didn't the author hypothesize that Goliath's interleaving of rows would degrade inference quality until it had been fine-tuned?
It will be interesting to see if this fine-tune supports that hypothesis.
Waiting for GUFF.