Installation

pip install git+https://github.com/koo-ec/HIP_HOPS_LLM.git

Python 3.10 or newer.

Extras

pip install "HIP-HOPS-LLM[all] @ git+https://github.com/koo-ec/HIP_HOPS_LLM.git"

Extra

Adds

For

bayes

pyagrum

Junction-tree inference and pyAgrum drawing. Optional — every quantity is also computable by exact enumeration in NumPy.

graph

langgraph, langchain-core

Reading a live compiled LangGraph. Not needed for mermaid text or specification dicts.

test

pytest, pyagrum, nbformat

Running the suite.

docs

sphinx, furo, myst-parser, …

Building these pages.

live

openai, anthropic, datasets

HIP-LLM’s live-provider paths.

all

everything above

The core needs only NumPy, pandas, matplotlib, SciPy and PyYAML.

Graphviz

pyAgrum draws through Graphviz, which is a separate native binary, not a Python package. pip install pydot does not provide it.

from hiphopsllm import graphviz_available

graphviz_available()      # runs `dot -V`, so it tells you the truth

If it returns False, nothing breaks: bn.show() draws the same network with matplotlib and says so in the caption. Install Graphviz only if you want pyAgrum’s richer rendering.

  • Ubuntu/Debian: sudo apt-get install graphviz

  • macOS: brew install graphviz

  • Windows: winget install graphviz (then reopen your terminal)

  • Conda: conda install -c conda-forge graphviz

Development install

git clone https://github.com/koo-ec/HIP_HOPS_LLM.git
cd HIP_HOPS_LLM
pip install -e ".[all]"
pytest

The suite includes HIP-LLM’s own tests, run against the vendored copy — see Vendoring.

Colab

!git clone --depth=1 https://github.com/koo-ec/HIP_HOPS_LLM.git hiphopsllm-repo
%pip install -q -e "hiphopsllm-repo[bayes,graph]"

Warning

Do not let the clone directory share a name with an importable module. In a notebook the working directory precedes site-packages on sys.path, so a directory called hiphopsllm (or HIP_HOPS_LLM, if you use the alias) is found first and imported as an empty namespace package. The import then succeeds and the module has no attributes at all:

AttributeError: module 'HIP_HOPS_LLM' has no attribute '__version__'

Cloning into hiphopsllm-repo, as above, avoids it entirely.

[all] also works, but it pulls the docs and release toolchains into the runtime, and on Colab that upgrades rich past what some preinstalled packages pin. [bayes,graph] is everything the notebook actually uses.

notebooks/HIP_HOPS_LLM_Colab.ipynb in the repository does this and runs the whole pipeline; open it directly from GitHub in Colab.