Gemma 7B in Oobabooga WebUI - Beats Mistral 7B and Llama2 13B Benchmarks?
HTML-код
- Опубликовано: 12 май 2024
- Gemma is a family of lightweight, state-of-the-art AI text generation models built by google from the same research and technology used to create their Gemini models, and it runs great in the Oobabooga text generation webui.
Being lightweight it will run on consumer hardware, and it's benchmark results beat other popular models such as Llama2 13B and Mistral 7B!
Come and chat to Gemma!
🙃Want to support the channel?
/ nerdyrodent
== Links ==
huggingface.co/google/gemma-7b
github.com/google/gemma_pytorch
github.com/oobabooga/text-gen...
github.com/ollama/ollama
== Stable Diffusion Links! ==
* Faster Stable Diffusions with the LCM LoRA - • LCM LoRA = Speedy Stab...
* How do I create an animated SD avatar? - • Create your own animat...
* Installing Anaconda for MS Windows Beginners - • Anaconda - Python Inst...
* Video-to-Video AI using AnimateDiff - • How To Use AnimateDiff...
* One image Gets You a Consistent Character in ANY pose - • Reposer = Consistent S... 
Наука
Maybe you'll have better luck with it than Matt Berman did, haha.
On a side note, thanks for the videos you make Nerdy, your videos are part of the reason I work in this industry now =) Seriously, you rock man!
Awesome!
@SoCalGuitarist How did you namage to find a job? I am trying to get into the industry as a full stack python dev
from the same research and technology Google used to create the Gemini models 🚩🚩🚩
Does it remove Europeans from their own history, too? I think I'll skip Google's comically political products.
6:18 If you consider "creamy" to be one syllable then it is indeed a haiku of 17 syllables in the proper format of 5-7-5
is it only working on english language or it can be used with any ?
One thing I'm not sure of is if these models can be fine tuned or merged with other models.
Oh, they have loads of info on fine tuning. I’m sure we’ll see some good ones pretty soon!
Cool video 😊😊😊. Much appreciated 😊
From what I know it's very low on accuracy considering it's 16bit. Some say that evaluations were cherry picked, plus it performs awfully using llama.cpp, in tests gemma.cpp should be used.
Will this model work with NVIDIA RTX chat?
Yea if you wonna use ray tracing and ambient occlusion
we appreciate you good sir. It's always a bright day to hear you giving us new things.
Thanks Nerdy, can I run this LLM on Chat With RTX?
Эта модель с цензурой?
100%
Not particularly, I’ve had it swear at me a lot!
Seems to fall into a repetition loop on the philosophical use of language being an everchanging tool.
Wonder if 7b models are more prone to this.
Worrying that it occurred within a few chat replies.
Yes, it’s a “weird” base model. Some of the fine tunes are pretty good though
🧀
lobotomized???
Yes but how racist is it?
There is no scientific basis for races.....
😊
@@MilesBellas The answer to nature or nurture is "it's complicated".
It's a known unknown.
/science class
@@MilesBellas then there can be no racism anywhere. Nice.
@@stefanb.8462
Races = Anachronistic = Unscientific
It's an arbitrary, outdated form of classification.
@@jtjames79
?
The comment refers to an anachronistic form of human classification... nothing else......
Maybe research?
Reading/Comprehension/Education
is this one woke too?
Rodents are better than Kittens ?!?!? I say fake news 🙄
Woke Gemma
Trust me, it's not beating Mistral at all. Gemma is considered as very very bad language model. You can't say it's better by asking " What is Red ". Please my friend, even you're Rock, Paper, Scisors was a failed and every 7b model can make it easily.