Prompt template for GPT4All:
<|endoftext|>Instruct: %1<|endoftext|> <|endoftext|>Output: %2<|endoftext|>
record low training loss on netcat420/MFANN for the 2.78b model!
(0.538)
previous record: (0.632)