Spaces:
Running
Running
update description and hf space dependency
Browse files- requirements.txt +1 -0
- serve/leaderboard.py +4 -2
requirements.txt
CHANGED
@@ -6,6 +6,7 @@ ase==3.23.0
|
|
6 |
torch==2.2.1
|
7 |
pymatgen==2024.4.13
|
8 |
bokeh
|
|
|
9 |
statsmodels==0.14.2
|
10 |
# py3Dmol==2.0.0.post2
|
11 |
# stmol==0.0.9
|
|
|
6 |
torch==2.2.1
|
7 |
pymatgen==2024.4.13
|
8 |
bokeh
|
9 |
+
bokeh_sampledata
|
10 |
statsmodels==0.14.2
|
11 |
# py3Dmol==2.0.0.post2
|
12 |
# stmol==0.0.9
|
serve/leaderboard.py
CHANGED
@@ -57,10 +57,12 @@ st.markdown(
|
|
57 |
"""
|
58 |
<h1 style='text-align: center;'>⚔️ MLIP Arena Leaderboard ⚔️</h1>
|
59 |
|
60 |
-
MLIP Arena is a platform for benchmarking foundation machine learning interatomic potentials (MLIPs).
|
61 |
-
""", unsafe_allow_html=True)
|
62 |
|
|
|
|
|
63 |
|
|
|
64 |
|
65 |
st.dataframe(
|
66 |
s,
|
|
|
57 |
"""
|
58 |
<h1 style='text-align: center;'>⚔️ MLIP Arena Leaderboard ⚔️</h1>
|
59 |
|
60 |
+
MLIP Arena is a platform for benchmarking foundation machine learning interatomic potentials (MLIPs), mainly for disclosing the learned physics and chemistry of the models and their performance on molecular dynamics (MD) simulations.
|
|
|
61 |
|
62 |
+
The benchmarks are NOT designed to compare model architectures, but to evaluate the readiness and reliability of the open-source, open-weight models to reproduce the qualitatively or quantitatively correct physics.
|
63 |
+
""", unsafe_allow_html=True)
|
64 |
|
65 |
+
st.header("Summary", divider=True)
|
66 |
|
67 |
st.dataframe(
|
68 |
s,
|