Alternate quantizations.
These are my own quantizations (updated almost daily).
The difference with normal quantizations is that I quantize the output and embed tensors to f16.
and the other tensors to 15_k,q6_k or q8_0.
This creates models that are little or not degraded at all and have a smaller size.
They run at about 3-6 t/sec on CPU only using llama.cpp
And obviously faster on computers with potent GPUs
More models here: https://huggingface.co/RobertSinclair
Indeed, it's much faster, just like driving a sports car with the same configuration, very good
Such technology deserves to be known and used by more people, so I won't be shutting it down
Indeed, it's much faster, just like driving a sports car with the same configuration, very good
Thanks. I strive to look where other people usually don't; it is second nature for me :D
Feel free to compare the various versions and tell me which ones is the best.
google/gemma-2-9b-it
google/gemma-2-9b-it
Thank you very much.gemma-2-9b-it sports car.
Request: aifeifei798/llama3-8B-DarkIdol-2.2-Uncensored-32K
Model name:
llama3-8B-DarkIdol-2.2-Uncensored-1048K
Model link:
https://huggingface.co/aifeifei798/llama3-8B-DarkIdol-2.2-Uncensored-1048K
Brief description:
The module combination has been readjusted to better fulfill various roles and has been adapted for mobile phones.
Uncensored 1048K
An image/direct image link to represent the model (square shaped):
Thank you very much
Request: aifeifei798/llama3-8B-DarkIdol-2.2-Uncensored-32K
Model name:
llama3-8B-DarkIdol-2.2-Uncensored-1048KModel link:
https://huggingface.co/aifeifei798/llama3-8B-DarkIdol-2.2-Uncensored-1048KBrief description:
The module combination has been readjusted to better fulfill various roles and has been adapted for mobile phones.
Uncensored 1048KAn image/direct image link to represent the model (square shaped):
Thank you very much
You're welcome.. quantizing it now.. p.s. what did you use to generate that gorgeous image?
Fooocus + animagine-xl-3.1.safetensors
Prompt:
wet,full body,sexy 18 yo Japanese Real girl in real, smiling at the camera,fashion glasses,gigantic breasts,poses,look at camera,llama3-8B-DarkIdol-2.2-Uncensored-1048K
Instructions: 50 were generated at once, then selected.
p.s. you should try uncensoring and training phi-3 small and phi-3 medium
Prompt:
wet,full body,sexy 18 yo Japanese Real girl in real, smiling at the camera,fashion glasses,gigantic breasts,poses,look at camera,llama3-8B-DarkIdol-2.2-Uncensored-1048K
Where to do that online? I could not find it in civit.ai
I didn't check their permission, is it okay to do this? phi-3 small and phi-3 medium Uncensored,
add me on discrod and let's talk
robert_46007
Request: aifeifei798/Phi-3-song-lyrics-1.0
Model name:
aifeifei798/Phi-3-song-lyrics-1.0
Model link:
https://huggingface.co/aifeifei798/Phi-3-song-lyrics-1.0
Brief description:
This model is a specialized model, specifically designed for writing song lyrics.
An image/direct image link to represent the model (square shaped):
WoW, I love girls shaped as the number 8 :D
perhaps a rounder face and chin would have been nicer.
Here, a sharp chin is considered a beauty standard. :p
Request: aifeifei798/llama3-8B-DarkIdol-2.3-Uncensored-32K
Model name:
aifeifei798/llama3-8B-DarkIdol-2.3-Uncensored-32K
Model link:
https://huggingface.co/aifeifei798/llama3-8B-DarkIdol-2.3-Uncensored-32K
Brief description:
The module combination has been readjusted to better fulfill various roles and has been adapted for mobile phones.
- Saving money(LLama 3)
- only test en.
- Input Models input text only. Output Models generate text and code only.
- Uncensored
- Quick response
- The underlying model used is winglian/Llama-3-8b-64k-PoSE (The theoretical support is 64k, but I have only tested up to 32k. :)
- A scholarly response akin to a thesis.(I tend to write songs extensively, to the point where one song almost becomes as detailed as a thesis. :)
- DarkIdol:Roles that you can imagine and those that you cannot imagine.
- Roleplay
- Specialized in various role-playing scenarios
An image/direct image link to represent the model (square shaped):