“decent speed” depends on your subjective opinion and what you want it to do. I think its fair to say if it can generate text around your slowest tolerable reading speed thats a bare minimum for real time conversational things. If you want a task done and don’t mind stepping away to get a coffee it can be much slower.
I was pleasantly suprised to get anything at all working on an old laptop. When thinking of AI my mind imagines super computers and thousand dollar rigs and data centers. I don’t think mobile computers like my thinkpad. But sure enough the technology is there and your old POS can adopt a powerful new tool if you have realistic expectations on matching model capacity with specs.
Tiny llama will work on a smartphone but its dumb. llama3.1 8B is very good and will work on modest hardware but you may have to be patient with it if especially if your laptop wasn’t top of the line when it was made 10 years ago. Then theres all the models in between.
The i7 duo core 2.6ghz CPU in my laptop trying to run 8B was jusst barely enough to be passing grade for real time talking needs at 1.2-1.7 T/s it could say a short word or half of a complex one per second. When it needed to process something or recalculate context it took a hot minute or two.
That got kind of annoying if you were getting into what its saying. Bumping the PC up to a AMD ryzen 5 2600 6 core CPU was a night and day difference. It spits out a sentence very quick faster than my average reading speed at 5-6 t/s. Im still working on getting the 4GB RX 580 GPU used for offloading so those numbers are just with the CPU bump. RAM also matters DDR6 will beat DDR4 speed wise.
Heres a tip, most software has the models default context size set at 512, 2048, or 4092. Part of what makes llama 3.1 so special is that it was trained with 128k context so bump that up to 131072 in the settings so it isnt recalculating context every few minutes…
Heres a tip, most software has the models default context size set at 512, 2048, or 4092. Part of what makes llama 3.1 so special is that it was trained with 128k context so bump that up to 131072 in the settings so it isnt recalculating context every few minutes…
Some caveats, this massively increases memory usage (unless you quantize the cache with FA) and it also massively slows down CPU generation once the context gets long.
TBH you just need to not keep a long chat history unless you need it,.
Another thing I’d recommend is running kobold.cpp instead of ollama if you want to get into the nitty gritty of llms. Its more customizable and (ultimately) faster on more hardware.
Thats good info for low spec laptops. Thanks for the software recommendation. Need to do some more research on the model you suggested. I think you confused me for the other guy though. Im currently working with a six core ryzen 2600 CPU and a RX 580 GPU. edit- no worries we are good it was still great info for the thinkpad users!
Yeah you should get kobold.cpp’s rocm fork working if you can manage it, otherwise use their vulkan build.
llama 8b at shorter context is probably good for your machine, as it can fit on the 8GB GPU at shorter context, or at least be partially offloaded if its a 4GB one.
I wouldn’t recommend deepseek for your machine. It’s a better fit for older CPUs, as it’s not as smart as llama 8B, and its bigger than llama 8B, but it just runs super fast because its an MoE.
Oh I got you mixed up with the other commenter, apologies.
I’m not sure when llama 8b starts to degrade at long context, but I wanna say its well before 128K, and where other “long context” models start to look much more attractive depending on the task. Right now I am testing Amazon’s mistral finetune, and it seems to be much better than Nemo or llama 3.1 out there.
“decent speed” depends on your subjective opinion and what you want it to do. I think its fair to say if it can generate text around your slowest tolerable reading speed thats a bare minimum for real time conversational things. If you want a task done and don’t mind stepping away to get a coffee it can be much slower.
I was pleasantly suprised to get anything at all working on an old laptop. When thinking of AI my mind imagines super computers and thousand dollar rigs and data centers. I don’t think mobile computers like my thinkpad. But sure enough the technology is there and your old POS can adopt a powerful new tool if you have realistic expectations on matching model capacity with specs.
Tiny llama will work on a smartphone but its dumb. llama3.1 8B is very good and will work on modest hardware but you may have to be patient with it if especially if your laptop wasn’t top of the line when it was made 10 years ago. Then theres all the models in between.
The i7 duo core 2.6ghz CPU in my laptop trying to run 8B was jusst barely enough to be passing grade for real time talking needs at 1.2-1.7 T/s it could say a short word or half of a complex one per second. When it needed to process something or recalculate context it took a hot minute or two.
That got kind of annoying if you were getting into what its saying. Bumping the PC up to a AMD ryzen 5 2600 6 core CPU was a night and day difference. It spits out a sentence very quick faster than my average reading speed at 5-6 t/s. Im still working on getting the 4GB RX 580 GPU used for offloading so those numbers are just with the CPU bump. RAM also matters DDR6 will beat DDR4 speed wise.
Heres a tip, most software has the models default context size set at 512, 2048, or 4092. Part of what makes llama 3.1 so special is that it was trained with 128k context so bump that up to 131072 in the settings so it isnt recalculating context every few minutes…
Some caveats, this massively increases memory usage (unless you quantize the cache with FA) and it also massively slows down CPU generation once the context gets long.
TBH you just need to not keep a long chat history unless you need it,.
Thank you thats useful to know. In your opinion what context size is the sweet spot for llama 3.1 8B and similar models?
Honestly as small as you can manage.
Again, you will get much better speeds out of “extreme” MoE models like deepseek chat lite: https://huggingface.co/YorkieOH10/DeepSeek-V2-Lite-Chat-Q4_K_M-GGUF/tree/main
Another thing I’d recommend is running kobold.cpp instead of ollama if you want to get into the nitty gritty of llms. Its more customizable and (ultimately) faster on more hardware.
Thats good info for low spec laptops. Thanks for the software recommendation. Need to do some more research on the model you suggested. I think you confused me for the other guy though. Im currently working with a six core ryzen 2600 CPU and a RX 580 GPU. edit- no worries we are good it was still great info for the thinkpad users!
8GB or 4GB?
Yeah you should get kobold.cpp’s rocm fork working if you can manage it, otherwise use their vulkan build.
llama 8b at shorter context is probably good for your machine, as it can fit on the 8GB GPU at shorter context, or at least be partially offloaded if its a 4GB one.
I wouldn’t recommend deepseek for your machine. It’s a better fit for older CPUs, as it’s not as smart as llama 8B, and its bigger than llama 8B, but it just runs super fast because its an MoE.
Oh I got you mixed up with the other commenter, apologies.
I’m not sure when llama 8b starts to degrade at long context, but I wanna say its well before 128K, and where other “long context” models start to look much more attractive depending on the task. Right now I am testing Amazon’s mistral finetune, and it seems to be much better than Nemo or llama 3.1 out there.