simle

app.py

@@ -19,16 +19,36 @@ from rdkit.Chem import DataStructs
 
 sys.path.insert(0, os.path.abspath("src/"))
 
-from clip.clip import _transform
-from training.datasets import CellPainting
-from clip.model import convert_weights, CLIPGeneral
-
 st.set_page_config(layout="wide")
 
 basepath = os.path.dirname(__file__)
 datapath = os.path.join(basepath, "data")
 
-MODEL_PATH = os.path.join(datapath, "epoch_55.pt")
-
 device = "cuda" if torch.cuda.is_available() else "cpu"
-model_type = "RN50"
+
+st.title('HyperDTI: Task-conditioned modeling of drug-target interactions.')
+
+def about_page():
+    st.markdown(
+        """        
+        HyperNetworks have been established as an effective technique to achieve fast adaptation of parameters for 
+        neural networks. Recently, HyperNetwork predictions conditioned on descriptors of tasks have improved 
+        multi-task generalization in various domains, such as personalized federated learning and neural architecture 
+        search. Especially powerful results were achieved in few- and zero-shot settings, attributed to the increased 
+        information sharing by the HyperNetwork. With the rise of new diseases fast discovery of drugs is needed which 
+        requires models that are able to generalize drug-target interaction predictions in low-data scenarios. 
+        
+        In this work, we propose the HyperPCM model, a task-conditioned HyperNetwork approach for the problem of 
+        predicting drug-target interactions in drug discovery. Our model learns to generate a QSAR model specialized on
+        a given protein target. We demonstrate state-of-the-art performance over previous methods on multiple 
+        well-known benchmarks, particularly in zero-shot settings for unseen protein targets.
+        """
+    )
+
+page_names_to_func = {
+    'About': about_page
+}
+
+selected_page = st.sidebar.selectbox('Choose function', page_names_to_func)
+
+page_names_to_func[selected_page]()