Install Ollama, your own personal LLM, on your Mac
ฝัง
- เผยแพร่เมื่อ 17 ต.ค. 2024
- In this video, I'm joined by José Domingo Cruz (the Mac Whisperer) as he helps me troubleshoot getting Homebrew, Docker, and Docker-Compose installed on his Mac. This makes it possible to use a docker-compose.yaml file to run Ollama with Open-Webui locally on a Mac.
Here are the basic steps and commands:
1. Install Homebrew on the Mac
/bin/bash -c "$(curl -fsSL raw.githubuser...)"
then later see "07:20 Finishing the Homebrew installation" for some extra commands to be copied and run from inside the terminal.
2. Install "Docker Desktop for Mac" from docs.docker.co... (NOTE: there are 2 versions: one for Apple silicon and one for Intel chips. Choose the one for your Mac)
3. Use Homebrew to install docker-compose
brew update
brew install docker-compose
4. Download the docker-compose.yaml file from Github (repo github.com/ada... )
raw.githubuser... (raw file)
5. In the same directory as the docker-compose file run docker compose to start the server
docker-compose up -d
(NOTE: to take the server offline use 'docker-compose down')
6. Visit localhost and download some models to start using your new local Ollama server.
----------------
Many thanks to the Mac Whisperer, José Domingo Cruz for his help in creating this video. I was lost without you buddy! You can find more of José at his website goldfish365.com/ . Also, if you're an English teacher looking for a way to get your students speaking fluently in the classroom, check his "Verbal Classrooms" page here goldfish365.co... !!!
Timestamps:
For the first three minutes we chat about what we're trying to do... (Hey, it was our first time ever doing this)
03:05 Getting started installing Homebrew
07:20 Finishing the Homebrew installation
08:45 Running brew update and installing docker-compose
09:38 YeeHa!
09:55 Checking that docker-compose is actually installed
11:05 Getting the docker-compose.yaml file (we actually use nano here to copy, paste and save - you could just download the file)
13:20 First attempt at running "docker-compose up -d"
14:00 Troubleshooting the first error message (no docker daemon)
17:50 Second attempt at running "docker-compose up -d"
18:20 Deciding to install "Docker Desktop for Mac"
19:40 Installing "Docker Desktop for Mac"
21:00 "Docker Desktop for Mac" first launch
22:35 Going through the first run licensing stuff
27:50 Third attempt at running "docker-compose up -d"
24:00 Troubleshooting the new error message (must use ASL logging - the answer was to run the command without "sudo")
25:20 Finding the answer
27:50 Fourth attempt at running "docker-compose up -d" (SUCCESS!!!)
28:15 Opening Ollama in a web browser for the first time and making a user account
29:45 Getting ready to add a model to the new Ollama
30:19 Checking the connection to the local Ollama API
30:48 Adding a model (mistral) to the new Ollama
33:01 Running the first prompt
34:30 Finding more models to download and information about each model
44:05 How to take the server down with 'docker-compose down'
Then we just chatted and had a look at the "Activity Monitor" while running a query etc. Basically we basked in the glory of success!
Links:
Ollama ollama.com
Open-WebUI github.com/ope...
Open-WebUI Community openwebui.com (modelfiles, prompts and more)
This is fantastic. I am not the most programming savvy guy and it ran perfectly in my computer. Thanks!
Great to hear! Enjoy your new AI!
This is awesome, thanks!
Glad you like it! Definitely felt a follow up to the workshop last Sunday was needed.