Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Next revision		 Previous revision
habrok:examples:ollama [2025/09/04 09:03] – created camarocicohabrok:examples:ollama [2025/10/15 11:42] (current) – pedro
Line 1: Line 1:
-{{indexmenu_n>4}} +====== Ollama on Hábrók ======
-===== Ollama on Habrok =====+
  
-You can run an LLM on Habrok with Ollama in a Jupyter environment by using the [[https://portal.hb.hpc.rug.nl/pun/sys/dashboard/batch_connect/sys/hb-ollama/session_contexts/new|Ollama (Jupyter)]] Interactive App on the [[https://portal.hb.hpc.rug.nl|Web Portal]].+You can run an LLM on Hábrók with Ollama in a Jupyter environment by using the [[https://portal.hb.hpc.rug.nl/pun/sys/dashboard/batch_connect/sys/hb-ollama/session_contexts/new|Ollama (Jupyter)]] Interactive App on the [[https://portal.hb.hpc.rug.nl|Web Portal]]. 
 +===== Setting up the virtual environment =====
  
-==== Setting up the virtual environment ==== +To be able to use the app, you first need to set up a Python virtual environment. The version of Ollama installed on Hábrók is ''ollama/0.6.0-GCCcore-12.3.0'', which means that the Python virtual environment needs to use a version of Python using the same ''GCCcore-12.3.0'' toolchain, which is ''Python/3.11.3-GCCcore-12.3.0''. Other versions of Python might work as well, but toolchain compatibilities can sometimes be an issue.
-To be able to use the app, you first need to set up a Python virtual environment. The version of Ollama installed on Habrok is ''ollama/0.6.0-GCCcore-12.3.0'', which means that the Python virtual environment needs to use a version of Python using the same ''GCCcore-12.3.0'' toolchain, which is ''Python/3.11.3-GCCcore-12.3.0''. Other versions of Python might work as well, but toolchain compatibilities can sometimes be an issue.+
  
-<code Bash>+<code shell>
 module load Python/3.11.3-GCCcore-12.3.0 module load Python/3.11.3-GCCcore-12.3.0
 python3 -m venv $HOME/venvs/ollama python3 -m venv $HOME/venvs/ollama
Line 14: Line 13:
 Once the virtual environment has been built, you need to install ''jupyter'' and ''ollama'' within the virtual environment; optionally, you can also install additional packages, such as ''openai'': Once the virtual environment has been built, you need to install ''jupyter'' and ''ollama'' within the virtual environment; optionally, you can also install additional packages, such as ''openai'':
  
-<code>+<code bash>
 source $HOME/venvs/ollama/bin/activate source $HOME/venvs/ollama/bin/activate
 pip install --upgrade pip pip install --upgrade pip
Line 21: Line 20:
  
 Finally, to make sure that the Jupyter Notebook is aware of your virtual environment, you need to create a Jupyter kernel: Finally, to make sure that the Jupyter Notebook is aware of your virtual environment, you need to create a Jupyter kernel:
-<code>+<code bash>
 python3 -m ipykernel install --user --name=ollama --display-name="Ollama" python3 -m ipykernel install --user --name=ollama --display-name="Ollama"
 </code> </code>
 +===== Choosing a folder for the models =====
  
-==== Choosing a folder for the models ==== 
 Another important choice when running the app is where the Ollama models should be saved; there are two options, with advantages and drawbacks: Another important choice when running the app is where the Ollama models should be saved; there are two options, with advantages and drawbacks:
   * **Custom directory**: This is a folder on the shared filesystem (we recommend ''/scratch/$USER'' since the models are quite large) where the models can be downloaded and saved for use in the future. This way, you only need to download a model once, but it will take quite a bit of time to save the model files to the shared filesystem. It might also be slower to use a model saved here.   * **Custom directory**: This is a folder on the shared filesystem (we recommend ''/scratch/$USER'' since the models are quite large) where the models can be downloaded and saved for use in the future. This way, you only need to download a model once, but it will take quite a bit of time to save the model files to the shared filesystem. It might also be slower to use a model saved here.
   * **Temporary directory**: This is a folder on the local disc of the node running the job, and is it considerably faster to save a downloaded model here; the drawback is that this is not persistent storage, and the model files will have to be downloaded for each session. However, it might also be a bit faster when using the model.   * **Temporary directory**: This is a folder on the local disc of the node running the job, and is it considerably faster to save a downloaded model here; the drawback is that this is not persistent storage, and the model files will have to be downloaded for each session. However, it might also be a bit faster when using the model.
 +===== Simple usage example =====
  
-==== Simple usage example ==== +To use Ollama in the Jupyter app, you need first to open a new notebook in the Jupyter app, and choose the **Ollama** Jupyter kernel built when setting up the virtual environment. Here is a small example which first imports the necessary packages: 
- +<code python>
-To use Ollama in the Jupyter app, here's a small example which first imports the necessary packages: +
- +
-<code Python>+
 import os import os
 import ollama import ollama
-    +
 from openai import OpenAI from openai import OpenAI
 </code> </code>
 then downloads a model from Ollama: then downloads a model from Ollama:
-<code Python>+<code python>
 ollama.pull("gemma3:12b") ollama.pull("gemma3:12b")
 </code> </code>
 and also lists all currently downloaded models: and also lists all currently downloaded models:
-<code Python>+<code python>
 for model in ollama.list().models: for model in ollama.list().models:
-  print(model.model)+    print(model.model)
 </code> </code>
 +
 It then creates a OpenAI API client: It then creates a OpenAI API client:
-<code Python>+<code python>
 client = OpenAI( client = OpenAI(
     base_url=f"http://{os.environ['OLLAMA_HOST']}/v1",     base_url=f"http://{os.environ['OLLAMA_HOST']}/v1",
Line 57: Line 55:
 </code> </code>
 and interacts with the LLM: and interacts with the LLM:
-<code Python>+<code python>
 response = client.chat.completions.create( response = client.chat.completions.create(
     model="gemma3:12b",     model="gemma3:12b",
Line 73: Line 71:
 print(response.choices[0].message.content) print(response.choices[0].message.content)
 </code> </code>
 +
 The model can, if desired, be deleted: The model can, if desired, be deleted:
-<code Python>+<code python>
 ollama.delete("gemma3:12b") ollama.delete("gemma3:12b")
 </code> </code>
  
 +You can find more info on how to use the Ollama Python library on their [[https://github.com/ollama/ollama-python|GitHub page]].