Train Your LLM with InstructLab v0.18.4: A Complete 4-GPU (L4s) Enterprise Hardware Walkthrough!

In this video, I show you how to install the required software on Red Hat Enterprise Linux 9.4 to enable GPU acceleration for InstructLab. We then install InstructLab v0.18.4 and enable cuda support. We add some spice by then installing vllm and enabling cuda support for ilab.
If you want to follow along, I am using a RHEL 9.4 box. Commands as shown in the video are as follows:
sudo dnf -y install cuda-toolkit-12-4
cd /usr/local
sudo rm cuda
sudo ln -s ./cuda-12.4 ./cuda
sudo dnf -y install libcudnn8 libcudnn8-devel cuda-cccl-12-4 libnccl-2.22.3-1+cuda12.4.x86_64
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/lib64
cd ~
mkdir instructlab
cd instructlab
python3.11 -m venv venv
source venv/bin/activate
rm -rf ~/.cache/pip
pip install instructlab
pip cache remove llama_cpp_python
pip install --force-reinstall "llama_cpp_python[server]==0.2.79" --config-settings cmake.args="-DLLAMA_CUDA=on"
pip install 'instructlab[cuda]'
pip install vllm@git+github.com/ope...
Clone the github.com/gsh... repo
ilab config init --train-profile PATH_TO_grantprofile.yaml from the above repo
Place taxonomy file (qna.yaml from above repo) into dir: ~/.local/share/instructlab/taxonomy/knowledge/time_travel
ilab taxonomy diff
ilab model download --repository TheBloke/Mixtral-8x7B-Instruct-v0.1-GPTQ --hf-token XXXXX
ilab model download --repository prometheus-eval/prometheus-7b-v2.0 --hf-token XXXXXXX
ilab model download --repository instructlab/granite-7b-lab
ilab data generate --model ~/.cache/instructlab/models/TheBloke/Mixtral-8x7B-Instruct-v0.1-GPTQ --gpus 4 --pipeline full
ilab model train --model-path instructlab/granite-7b-lab --data-path ~/.local/share/instructlab/datasets/knowledge_train_msgs….jsonl