Note that this is StableLM ALPHA (only 0.52 epochs into training). The fully trained version will surely be much better. Also, you should benchmark GPT-3 Babb

Note that this is StableLM ALPHA (only 0.52 epochs into training). The fully trained version will surely be much better. Also, you should benchmark GPT-3 Babbage for a fair comparison since that is the same size as 7B.