| --- | |
| library_name: flax | |
| tags: | |
| - chemistry | |
| - ionic-conductivity | |
| - polymer-electrolytes | |
| - jax | |
| pipeline_tag: tabular-regression | |
| --- | |
| # Arrhenius Predictor for Polymer Electrolytes | |
| Paper: https://pubs.acs.org/doi/10.1021/acscentsci.2c01123 | |
| This model predicts ionic conductivity ($\ln \sigma$), activation energy ($E_a$), and the Arrhenius prefactor ($\ln A$) for polymer electrolytes. It uses a physics-informed architecture where the output is constrained by the Arrhenius equation: | |
| $$ \ln \sigma = \ln A - \frac{E_a}{RT} $$ | |
| ## Usage | |
| This model expects inputs processed via `MolGraphizer` and `expand_polymer_smiles`. | |
| It requires the following features: | |
| - **Molecular Graph**: Nodes, Edges, Connectivity. | |
| - **Auxiliary Features**: Temperature (K), Log Molecular Weight, Molality. | |
| To use this model for screening new candidates, use the `screen_from_hub.py` script provided in the repository. | |
| ```python | |
| # Pseudo-code for loading | |
| import flax.serialization | |
| from huggingface_hub import hf_hub_download | |
| path = hf_hub_download(repo_id="eamag/chemarr", filename="model.msgpack") | |
| with open(path, "rb") as f: | |
| artifact = flax.serialization.from_bytes(dummy_state, f.read()) | |