add hf pretrained encoders

Author: liugangcode

docs/source/api/encoder.rst

@@ -74,4 +74,13 @@ Supervised Pretraining for Molecules
    :members: fit, encode
    :exclude-members: fitting_epoch, fitting_loss, model_name, model_class
    :undoc-members:
-   :show-inheritance:
+   :show-inheritance:
+
+Pretrained Molecular Encoders
+----------------------------------------------
+
+.. rubric:: Sequence-based Pretrained Transformers from Hugging Face
+
+.. autoclass:: torch_molecule.encoder.pretrained.modeling_pretrained.HFPretrainedMolecularEncoder
+   :members: fit, encode, load, load_from_hf
+   :exclude-members: fitting_epoch, fitting_loss, model_name, model_class

torch_molecule/encoder/pretrained/modeling_pretrained.py

@@ -31,7 +31,7 @@ class HFPretrainedMolecularEncoder(BaseMolecularEncoder):
     """Implements Hugging Face pretrained transformer models as molecular encoders.
 
     This class provides an interface to use pretrained transformer models from Hugging Face
-    as molecular encoders. It handles tokenization, encoding, and pooling of molecular representations.
+    as molecular encoders. It handles tokenization and encoding of molecular representations.
 
     Tested models include:
 
@@ -66,6 +66,8 @@ class HFPretrainedMolecularEncoder(BaseMolecularEncoder):
       Output dimension: 384.
       repo_id: "seyonec/ChemBERTa-zinc-base-v1" (https://huggingface.co/seyonec/ChemBERTa-zinc-base-v1)
 
+    Other models accessible through the transformers library have not been explicitly tested but may still be compatible with this interface.
+
     Parameters
     ----------
     repo_id : str
@@ -140,9 +142,11 @@ def save_to_hf(self) -> None:
         raise NotImplementedError("PretrainedMolecularEncoder does not support saving to huggingface.")
 
     def load_from_hf(self) -> None:
+        """The same as fit()"""
         self.fit()
 
     def load(self) -> None:
+        """The same as fit()"""
         self.fit()
 
     def fit(self) -> "HFPretrainedMolecularEncoder":