106 lines
3.1 KiB
Markdown
106 lines
3.1 KiB
Markdown
# Install Ollama
|
|
|
|
```
|
|
curl -fsSL https://ollama.com/install.sh | sh
|
|
```
|
|
|
|
I prefer to make the model files available for all computers in our working group. Thus I put them on a NAS under /data_1/deepseek.
|
|
|
|
|
|
In that case you want to change /etc/systemd/system/ollama.service and add your directory to it:
|
|
```
|
|
Environment= [PATH BLA BLA] "OLLAMA_MODELS=/data_1/deepseek/models" "HOME=/data_1/deepseek"
|
|
```
|
|
|
|
After that restart systemd with:
|
|
```
|
|
systemctl restart ollama.service
|
|
```
|
|
|
|
Check the status of the service:
|
|
```
|
|
systemctl status ollama.service
|
|
```
|
|
## Getting the models
|
|
|
|
Now we can get the models:
|
|
```
|
|
ollama pull deepseek-r1:1.5b
|
|
ollama pull deepseek-r1:7b
|
|
ollama pull deepseek-r1:8b
|
|
ollama pull deepseek-r1:32b
|
|
ollama pull deepseek-r1:14b
|
|
ollama pull deepseek-r1:70b
|
|
```
|
|
|
|
However, you want to check first which model you want based on the CPU RAM or the GPU VRAM you have:
|
|
|
|
```
|
|
ollama list | grep deepseek
|
|
```
|
|
|
|
```
|
|
deepseek-r1:1.5b a42b25d8c10a 1.1 GB
|
|
deepseek-r1:671b 739e1b229ad7 404 GB
|
|
deepseek-r1:32b 38056bbcbb2d 19 GB
|
|
deepseek-r1:14b ea35dfe18182 9.0 GB
|
|
deepseek-r1:7b 0a8c26691023 4.7 GB
|
|
deepseek-r1:70b 0c1615a8ca32 42 GB
|
|
deepseek-r1:8b 28f8fd6cdc67 4.9 GB
|
|
```
|
|
## Test
|
|
|
|
```
|
|
ollama run deepseek-r1:1.5b
|
|
```
|
|
|
|
```
|
|
>>> Hello
|
|
<think>
|
|
|
|
</think>
|
|
|
|
Hello! How can I assist you today? 😊
|
|
|
|
>>> /bye
|
|
```
|
|
|
|
# Using it with VS Code
|
|
|
|
For using Ollama we need a special setting for VS Code. Thus we need to produce instances with different model parameter (or in other words: got to the Modelfile subfolder and check the information their)
|
|
```
|
|
code_ds32b:latest 995e2d04e071 19 GB
|
|
code_ds70b:latest 4930f987452d 42 GB
|
|
code_ds7b:latest 0438bd669fa8 4.7 GB
|
|
code_ds8b:latest 643346a4074c 4.9 GB
|
|
code_ds1.5b:latest 2d66604e7b60 1.1 GB
|
|
code_ds14b:latest 76c930e3d70a 9.0 GB
|
|
```
|
|
Do not you larger models on your CPU or you will die of old age!
|
|
|
|
## Install roo code
|
|
|
|
Please make sure that your shell doesn't use something like Starship or the posh packages! Otherwise VS Code can not run terminal command!
|
|
|
|
|
|
|
|
## Setting of roo code
|
|
|
|
You need to select Ollama. Then you need to connect to Ollama. If Ollama is local on your machine, you don't so anything here. For home office
|
|
I use
|
|
```
|
|
ssh -p [PORT OLLAMA SERVER] -L 11434:127.0.0.1:11434 [USERNAME]@[IP OLLAMA SERVER]
|
|
```
|
|
to connect directly to the computer with Ollama and tunnel the 11434 to my local machine. (Or in other words: I make the external Ollama available as local service on port 11434).
|
|
|
|
Select a code model, which you created by using the modelfiles.
|
|
|
|
|
|
|
|
## Trust issues
|
|
|
|
Now you need to decide how much trust you have and allow Roo Code to do stuff without asking you all the time.
|
|
|
|
And don't ask me what MCPs servers are. I don't know... yet.
|
|
|
|
|