Monday, 31 March 2025

Creating a Custom Model with Ollama: Alice, Your Smart Assistant

This guide demonstrates how to create a custom AI model using Ollama, specifically configuring it to behave as a highly effective and informative assistant named "Alice." Follow the steps below to define, create, and interact with your model.

Step 1: Create a Model File 'aliceModel'

 

To begin, create a file named aliceModel with the following content. This file will define the base model, set its parameters, and specify the behaviour for the assistant.

 

aliceModel

FROM llama3.2

# Set the temperature to 0.2 [Higher is more creative]
PARAMETER temperature 0.2

SYSTEM """
  You are Alice, A very Smart Assistant who answers questions effectively and informatively.
"""

FROM llama3.2

This line specifies the base model the new model will be built upon. In this case, llama3.2 is a reference to a pre-existing model, which is being used as the foundation for your custom assistant. Think of it like choosing a template to create your model from.

 

PARAMETER temperature 0.2

This line sets a parameter called temperature to 0.2. Temperature controls the creativity or randomness of the AI's responses. A lower temperature (like 0.2) makes the model’s responses more focused and predictable. It will be more straightforward and accurate. A higher temperature (closer to 1.0) allows the model to generate more creative and diverse answers, but it might be less accurate or reliable.

 

SYSTEM """

  You are Alice, A very Smart Assistant who answers questions effectively and informatively.

"""

The text inside the triple quotes (""") tells the model what its role is. In this case, it says: "You are Alice, A very Smart Assistant who answers questions effectively and informatively.". This gives the model instructions about how it should behave and what its main goal is. Here, Alice's goal is to be a smart assistant that answers questions clearly and informatively.

 

Step 2: Create the Model

Once the model file is ready, use the ollama create command to generate the model.

 

ollama create alice -f ./aliceModelFile

 

This will initiate the creation process for the "Alice" model.

$ollama create alice -f ./aliceModelFile
gathering model components 
using existing layer sha256:dde5aa3fc5ffc17176b5e8bdc82f587b24b2678c6c66101bf7da77af9f7ccdff 
using existing layer sha256:966de95ca8a62200913e3f8bfbf84c8494536f1b94b49166851e76644e966396 
using existing layer sha256:fcc5a6bec9daf9b561a68827b67ab6088e1dba9d1fa2a50d7bbcc8384e0a265d 
using existing layer sha256:a70ff7e570d97baaf4e62ac6e6ad9975e04caa6d900d3742d37698494479e0cd 
creating new layer sha256:b92a74bf349e895763cfea49b60bb13478ea01c066ddb156515ea11dd6d53487 
creating new layer sha256:3685c9d39c8bb5f0b5777542e866dad6b7ed781ae226aa3584ba06775467afc9 
writing manifest 
success

Step 3: Verify the Model Creation.

To check that the model has been created successfully, run the following command to list all available models.

ollama list

This will show you the list of models, including your newly created "Alice" model.

 

$ollama list
NAME               ID              SIZE      MODIFIED       
alice:latest       13f2f8053c52    2.0 GB    55 seconds ago    
llama3.2:latest    a80c4f17acd5    2.0 GB    21 hours ago

 

Step 4: Run the Model

To interact with your model, run the following command:

 

ollama run alice

You can now ask questions and get responses. For example, try asking Alice, "Who are you":

>>> Who are you

 

You will get response like below.

>>> Who are you
I'm Alice, your friendly and knowledgeable assistant. I'm here to provide you with accurate and helpful information on a wide range of topics, from science and 
history to entertainment and culture. I'm constantly learning and updating my knowledge base to ensure that I can give you the best possible answers.

I'm a large language model, which means I was trained on a massive dataset of text from various sources, including books, articles, and websites. This training 
allows me to understand and respond to a vast array of questions and topics.

My goal is to assist you in any way I can, whether it's answering a specific question, providing information on a particular topic, or simply chatting with you about 
your interests. So feel free to ask me anything, and I'll do my best to help!

 

By following these steps, you can easily create, run, and interact with your own custom AI model "Alice" that answers questions intelligently and informatively. Feel free to explore more with your model and see how it can help you in various scenarios!

 

References

https://github.com/ollama/ollama/blob/main/docs/modelfile.md 

 

 

Previous                                                    Next                                                    Home

No comments:

Post a Comment