Will there be a 33B version?

#1
by Izumicc - opened

It works perfect well, better than chinese-alpaca-plus-13b in my use case. Can I expect a 33B model?

Actually I merged a 33B version.However, due to the lack of compute resources, I only could load the 4bit version of 33B, which performs close to the 13B version. As there is a need for the 33B version, I shall upload them.

Thank you for your work! Actually I can only run the quantification models too(ggml format). 8_0 for 13B and 4_s for 33B, despite this the 33b is still much better than the 13B in most cases

Izumicc changed discussion status to closed

Sign up or log in to comment