Update README.md

README.md

@@ -1,3 +1,37 @@
 ---
 license: mit
+tags:
+- chemistry
 ---
+Here is the dataset used in paper `Scalable Machine Learning Force Fields for Macromolecular Systems Through Long-Range Aware Message Passing` [[link](URL地址)].
+
+The data is stored in lmdb formate. Taking `md_traj` data as an example, you can load the data as following:
+```python
+import lmdb
+import pickle
+env = lmdb.open("md_traj/train/NaCl/data_0.lmdb", subdir=False)
+txn = env.begin()
+length = txn.stat()['entries']
+data_list = []
+for idx in range(length):
+    data.append(pickle.loads(txn.get(f"idx".encode())))
+print(data_list[0].keys())
+```
+You will get:
+```python
+['forces',            # Forces in kcal/mol/Angstrom
+ 'cluster_ids',       # The id for each atoms in prebuilt clusters
+ 'order',             # The frame id in MD simulation
+ 'pos',               # The positons for each atoms, shape [N, 3], unit in Angstrom
+ 'cluster_centers',   # The center of prebuilt clusters
+ 'energy',            # The total energy of the molecule in kcal/mol
+ 'atomic_numbers']    # The list of atomic numbers, shape [N,]
+```
+For Di-Molecule dataset, there are additional properties:
+```python
+[...
+ 'mol_a',             # The pubchem id of the molecule A
+ 'mol_b',             # The pubchem id of the molecule B
+ 'distance',          # The horizontal distance between the right-most atom of A and the left-most atom of B, unit in Angstrom
+...]
+```