InstructLab Knowledgebase Guide
Overview
This guide explains how to set up, run, and use InstructLab, including what to do after running the setup script and after logging out and back in.
Activate the environment:
- Each time you log in, activate the virtual environment:
|
source ~/uv-instructlab/bin/activate
|
Initialize InstructLab (only needs to be done once per environment):
When prompted:
- Taxonomy repository: Press Enter to clone the default repository.
- Model path: Press Enter to use the default model, or provide a custom path (default model: Merlinite). See below for more details on models.
- System profile selection:
- If you are using a CAEN linux machine (remote or physical), when prompted:
- Select 3 for Intel.
- Select 1 for normal CPU-based system:
- Once you see the prompt below, you have finished the initialization process:
Download Model:
Even after initialization, you must download the default model using this command:
To check out different models available, see this page: https://docs.instructlab.ai/getting-started/download_models/
Running the Model
Once you see a line like this, you can chat with the model:
Keep this terminal open while using the model. To use a custom model (optional):
|
ilab model serve --model-path models/granite-7b-instruct.gguf
|
Chatting with the Model
- Open a new terminal, activate the environment, then run:
Exiting and stopping
- To exit the chatbot, type exit or quit, or press Ctrl + c
- To stop the server, press Ctrl + c in the terminal running “ilab model serve”
After Logging Out / Reconnecting
When you return:
- Activate the environment:
|
source ~/uv-instructlab/bin/activate
|
- Start the model:
- Open a new terminal, activate the environment, and chat:
Configuration Details
- To change the default model, open the config file in a text editor:
|
nano ~/.config/instructlab/config.yaml
|
|
model: path: /home/uniquname/.cache/instructlab/models/your-new-model.gguf
|
Additional Resources: