Start JupyterLab in HPC

hpcf_interactive

export PATH=$PATH:"/home/yli11/HemTools/bin"

module load python/2.7.13

run_lsf.py -p jupyterlab --memory 20000

You will receive an email with the link to the jupyter notebook within 5 minutes.

Start JupyterLab-AI in HPC (qwen3.5)

hpcf_interactive

export PATH=$PATH:"/home/yli11/HemTools/bin"

module load python/2.7.13

run_lsf.py -p jupyterai_qwen --memory 50000

Then go to jupyternaut setting, set openai/Qwen/Qwen3.5-27B-FP8

Note, using the chat box is faster

Open a jupyter notebook and set

%reload_ext jupyter_ai_magic_commands

%ai alias my_model openai/Qwen/Qwen3.5-27B-FP8

%config AiMagics.initial_language_model = "my_model"

%%ai -f code
write hello world

%%ai -f code
A function that computes the lowest common multiples of two integers, and
a function that runs 5 test cases of the lowest common multiple function

Start JupyterLab-AI in HPC (llama3.3-70b)

hpcf_interactive

export PATH=$PATH:"/home/yli11/HemTools/bin"

module load python/2.7.13

run_lsf.py -p run_jupyterai --memory 20000

You will receive an email with the link to the jupyter notebook within 5 minutes.

first go to jid folder tail litellm.log to find the port for AI server.

Then go to jupyternaut setting, set openai/gpt-5 and http://localhost:48219

Model used is llama3.3-70b-instruct-vllm, not gpt-5: gpt-oss-120b-16bit-vllm, because of speed. But I still name it as openai/gpt-5.

Open a jupyter notebook and set

%reload_ext jupyter_ai_magic_commands

%config AiMagics.initial_language_model = "openai/gpt-5"

%%ai -f code
A function that computes the lowest common multiples of two integers, and
a function that runs 5 test cases of the lowest common multiple function

You can also use the chat box

Set up env (my notes for installation)

module load conda3/202011
conda create -n captureC -c bioconda bioconductor-chicago
source activate captureC
conda install -c anaconda jupyter_client
conda install -c conda-forge jupyterlab
conda install -c r r-pbdzmq
conda install -c conda-forge nodejs
conda install -c cheng_lab easy_prime # to install some python libraries
jupyter labextension install @techrah/text-shortcuts
export XDG_RUNTIME_DIR=""
export JUPYTER_RUNTIME_DIR=/tmp
jupyter lab --ip='*' --NotebookApp.token='' --NotebookApp.password='' --no-browser --port 9865

in R

install.packages('IRkernel')
IRkernel::installspec()

OLD

Previously I have setup a code to run jupyter notebook as below:

export PATH=$PATH:"/home/yli11/HemTools/bin"

module load python/2.7.13

run_lsf.py -p jupyter --memory 20000

You will receive an email with the link to the jupyter notebook.

To use jupyter ai locally, I have modified

::

/home/yli11/.conda/envs/jupyter_ai/lib/python3.13/site-packages/jupyter_ai_jupyternaut/jupyternaut/jupyternaut.py # streaming to false aiohttp_transport.py # ssl verify to false chat_models.py # disable audio and catch token

code @ github.