How to make 8 bit sharded model?

#1
by a1nkit - opened

Can you plz share the steps involved in making 8 bit sharded model?

Analytics Club at ETH Zürich org

hi! Simply load the model in 8-bit with the bitsandbytes integration as described here and then you can shard/push it as I described in this discussion on another model.

hope that helps and if you have further issues/questions feel free to comment here or reopen as needed.

pszemraj changed discussion status to closed

Sign up or log in to comment