jan-janssen / LangSim

Application of Large Language Models (LLM) for computational materials science - visit jan-janssen.com/LangSim

Home Page:http://jan-janssen.com/LangSim/

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Install LangSim locally

jan-janssen opened this issue · comments

There are currently different ways to install LangSim locally:

git clone https://github.com/jan-janssen/LangSim
cd LangSim
conda env create -f environment.yml --name LangSim

Currently, the code requires pymace which is not included in the environment.yml, it can be installed using:

conda install -c conda-forge pymace

Afterwards you can either install the LangSim package locally:

pip install .

Alternatively, if you do not want to install the package you can either set the python path:

export PYTHONPATH="/path/to/LangSimRepository":$PYTHONPATH

Or you execute the tests directly in the repository folder, in that case you have to copy the demonstration.ipynb notebook from the notebooks folder:

mv notebooks/demonstration.ipynb . 

I encountered a minor issue that I don't remember seeing in the remote version:

langchain ValueError: An output parsing error occurred. In order to pass this error back to the agent and have it try again, pass `handle_parsing_errors=True` to the AgentExecutor. This is the error: Could not parse tool input: {'arguments': '{\n"atom_dict": {\n "numbers": [47],\n "positions": [[0.0, 0.0, 0.0]],\n "cell": [[0.0, 2.045, 2.045], [2.045, 0.0, 2.045], [2.045, 2.045, 0.0]],\n "pbc": [True, True, True]\n},\n"calculator_str": "emt"\n}', 'name': 'get_atom_dict_equilibrated_structure'} because the `arguments` is not valid JSON.

It's easy enough to fix by setting handle_parsing_errors=True when initializing the AgentExecutor in llm.py, line 70.

Does this happen with ChatGPT 4? I remember I had a similar issue with 4 but not with 4o.

I saw it with GPT 4, not 4o. Looks a bit non-specific.

The installation instructions are now also included in the website.