Matt, your approach in explaining how and why a PR can be done is concise, precise and very approachable. Apart from learning that contexts can be quantized, I really liked your way of explaining the PR process. Kudos!
Which one do you prefer for artificial intelligence? Mac Studio Apple M2 Ultra with 24‑core CPU, 76‑core GPU, 32‑core Neural Engine 192GB unified memory 8TB SSD storage Front: Two Thunderbolt 4 ports, one SDXC card slot Back: Four Thunderbolt 4 ports, two USB-A ports, one HDMI port, one 10Gb Ethernet port, one 3.5 mm headphone jack Final Cut Pro Logic Pro Or 2(Two) Nvidia 4090 ? Ignore the price difference. Just compare power and speed. Consider 3rd party hardware such as display, keyboard, mouse but ignore price. It's just a question. I'm not going to buy it right now. Thanks
If you are running Linux, this will stop the default server "sudo systemctl stop ollama". I know nothing about Macs but it might work there too since Macs run a quasi-unix like kernel.
I managed to install gcc after much searching. I installed go. An hour and a half has gone by. I type 'make -j 5' and course make is not recognized. I suspected that as soon as it became apparent that it wasn't a windows thing but more Linuxy there would be endless problems, and it wouldn't work. Linux never disappoints. I know, it's a 'dev' channel not for stupid hobbyists.
Hello and thank you I’ll try for sure I have an unrelated request We are trying to build a custom application on olllama with mistral (..being in France) The two end user interfaces that I prefer (page assist on chrome and open-Webui) are not ok because we do not want to give end users all those options So can you show us how to build a web app that has built in the parameters and the model and just have the Chat interface with upload? maybe some python app or maybe something that exist and then don’t know of? ) Thanks for all your videos
Matt can i do this in docker? or is there any docker image ready to use? so i just can use other port and make messy stuff to beta test those PR in docker? maybe that wa can save lots of time?
Sure you can and thats a good idea. But it gets in that "complex" field I think Matt is willing to avoid go through. But you may have a "base" docker Linux image with git, gh and GoLang installed, and run different containers with different Pool Requests as he explained.
Your videos have been fantastic in unlocking a few of the more esoteric aspects of what seems to be a simple software like ollama. Thanks for all of these videos. I am building my own SaaS company and some of these videos have really helped me shortcut the discovery and learning process.
if i have infinite life span i might try this.... however my life span spent for feeding me and my family i really have so many things i want to try to do but 24 hour isn't enough 100 years life span also isn't enough... i always wanted ollama can multimodal capability especially that streaming token features...
I did manage to install make. Last file revision 2006. So I input the make command from inside the directory and lo and behold, success. Hah, I am joking just another half a screen of errors. So, I follow instructions. Get difficult get requirements, do as instructed and get gibberish. That will be all from me. Can't even compile the main branch. Can't even get make to run. Funny, I can pull Unreal Engine from Github and load it into Visual Studio and compile it to get my very own build. No worries. But when it comes to things with a connection to Linux there is no hope. Unless one hopes for hundreds of hours of wasted time and enjoy reading pages of cryptic gibberish.
iv nvr usd pr b4 but +1 is clearly a vote up like readit whats your issue lol less is more conversely if you see -1 then its a bad but saying its bad without saying why its bad is more of an issue than +1 lol
@@technovangelist I think they're confusing pressing a like / upgoat button with people commenting on an issue. But at this point it's anyone's guess 😂
Thanks for featuring the K/V quant PR Matt, and importantly for promoting constructive PR behaviour.
Matt, your approach in explaining how and why a PR can be done is concise, precise and very approachable. Apart from learning that contexts can be quantized, I really liked your way of explaining the PR process. Kudos!
The most valuable and knowledgeable LLM trainer alive. Matt is priceless and his teaching style surpasses all trainers. Great job!💥
Thank you, the quantization for context is very useful for my projects 😊
Great explanation video! Love it!
A +1 to this video Matt 😉
Thank you for your new knowledge.
Good and clear information, keep it going.
Hola saludos desde el Perú 🇵🇪👍🏼
Which one do you prefer for artificial intelligence?
Mac Studio
Apple M2 Ultra with 24‑core CPU, 76‑core GPU, 32‑core Neural Engine
192GB unified memory
8TB SSD storage
Front: Two Thunderbolt 4 ports, one SDXC card slot
Back: Four Thunderbolt 4 ports, two USB-A ports, one HDMI port, one 10Gb Ethernet port, one 3.5 mm headphone jack
Final Cut Pro
Logic Pro
Or
2(Two) Nvidia 4090 ?
Ignore the price difference. Just compare power and speed. Consider 3rd party hardware such as display, keyboard, mouse but ignore price. It's just a question. I'm not going to buy it right now. Thanks
M4 max or wait for the m4 ultra
If you are running Linux, this will stop the default server "sudo systemctl stop ollama". I know nothing about Macs but it might work there too since Macs run a quasi-unix like kernel.
depending on your setup, you may or may not require sudo, but otherwise that is accurate for linux. It's different on mac.
I managed to install gcc after much searching. I installed go. An hour and a half has gone by. I type 'make -j 5' and course make is not recognized.
I suspected that as soon as it became apparent that it wasn't a windows thing but more Linuxy there would be endless problems, and it wouldn't work.
Linux never disappoints.
I know, it's a 'dev' channel not for stupid hobbyists.
I have a "mini-pc" in my homelab setup with iGPU. going to try and run Ollama on that one,. so i cna have 100% uptime on it :D
what is your igpu?
I also have mini pc with igpu
Hello and thank you I’ll try for sure
I have an unrelated request
We are trying to build a custom application on olllama with mistral (..being in France)
The two end user interfaces that I prefer (page assist on chrome and open-Webui) are not ok because we do not want to give end users all those options
So can you show us how to build a web app that has built in the parameters and the model and just have the Chat interface with upload? maybe some python app or maybe something that exist and then don’t know of? )
Thanks for all your videos
You got hit by the storms too ? I'm in northern whatcom and heard a bunch of houses were totalled nearby
Ps: trying out the kv build today
Yup. I’m on Bainbridge island
Hope you guys were ok!
Matt can i do this in docker? or is there any docker image ready to use? so i just can use other port and make messy stuff to beta test those PR in docker? maybe that wa can save lots of time?
Sure you can and thats a good idea. But it gets in that "complex" field I think Matt is willing to avoid go through. But you may have a "base" docker Linux image with git, gh and GoLang installed, and run different containers with different Pool Requests as he explained.
To use gcc on Windows, you should just install WSL for ease of use
I have WSL and tried gcc command in a windows terminal and it does not work
+1
Your videos have been fantastic in unlocking a few of the more esoteric aspects of what seems to be a simple software like ollama.
Thanks for all of these videos. I am building my own SaaS company and some of these videos have really helped me shortcut the discovery and learning process.
if i have infinite life span i might try this.... however my life span spent for feeding me and my family i really have so many things i want to try to do but 24 hour isn't enough 100 years life span also isn't enough...
i always wanted ollama can multimodal capability especially that streaming token features...
I did manage to install make. Last file revision 2006. So I input the make command from inside the directory and lo and behold, success. Hah, I am joking just another half a screen of errors. So, I follow instructions. Get difficult get requirements, do as instructed and get gibberish. That will be all from me. Can't even compile the main branch. Can't even get make to run.
Funny, I can pull Unreal Engine from Github and load it into Visual Studio and compile it to get my very own build. No worries.
But when it comes to things with a connection to Linux there is no hope. Unless one hopes for hundreds of hours of wasted time and enjoy reading pages of cryptic gibberish.
iv nvr usd pr b4 but +1 is clearly a vote up like readit whats your issue lol less is more conversely if you see -1 then its a bad but saying its bad without saying why its bad is more of an issue than +1 lol
I have no idea what you are trying to say.
@@technovangelist I think they're confusing pressing a like / upgoat button with people commenting on an issue. But at this point it's anyone's guess 😂