https://huggingface.co/Darkknight535/Moonlight-L3-15B-v2.5-64k
#371
by
Ttimofeyka
- opened
This is 15B 64k, which you have already quantized, but is more coherent because of the Lunaris merge. I hope the author will make a 512k version based on my 15B 512k Instruct, but I think it will spray the model too much.
512k :\
well, let's quant it :) as usual, progress can be followed at http://hf.tst.eu/status.html
mradermacher
changed discussion status to
closed
F mradermacher
very F. but after patching llama.cpp, it's now green F.
Ayo no way.. lol