Good model
L3 70B Abliterated v1 is my daily model for most tasks (along Miqu 70b and Kyllene 34b).
This daybreak version works as one would expect of such combo, without losing the qualities of the base model!
Thanks for the feedback! This is my go to model currently as well.
Hey!
Could you make and a F16 (or BF16) GGUF?
I want to make my own quants, and I am unable to make the conversion (I'm blocked at model-00013-of-00030.safetensors when I load the model, no matter which build of LCPP I'm using), even after redownloading the weight again. :/
I don't know what's wrong on my side.
Sure thing. For the record, are you using the --use-temp-file
flag? I will run out of RAM if I don't use that, unless I enable swap, and I have lots of RAM.
No, I didn't, I don't have this problem. I will make a swap of 128GB just to see.
(and thank you!)
Edit : It worked, thank you very much lol.
OK! My f16 just finished, but it looks like you don't need it anymore. Tell me if you do and I'll upload it.
Oh, please do.
If you did the f16, i'll upload the bf16!
Note : Finally, I will be able to dump the IQ3_M I was using for a better one. xD
All right, uploading to https://huggingface.co/crestf411/L3-70B-daybreak-abliterated-v0.4-f16-gguf now.
Btw did you try the sunfall model(s)? They are basically an extension of daybreak. Very experimental though.
I forgot about the file size cap. Splitting it now. My HDD is ultra slow for some reason though so it'll be a while before I upload it. Edit: it's up.