Update app.py

app.py

@@ -41,7 +41,7 @@ def create_model(optimizer='RMSprop', learn_rate=0.1, momentum=0.4, activation='
     keras_model.add(Dropout(dropout_rate))
     keras_model.add(Dense(1,activation='linear'))
     keras_model.summary()
-    # Compile model
+    
     keras_model.compile(loss='mean_squared_error', optimizer=optimizer)
 
     return keras_model
@@ -61,9 +61,9 @@ def get_ecfc(smiles_list, radius=2, nBits=2048, useCounts=True):
             else:    
                 ecfp_fingerprints.append(list(AllChem.GetMorganFingerprintAsBitVect(mol, radius, nBits).ToBitString()))  
     
-    # Create dataframe of fingerprints
+
     df_ecfp_fingerprints = pd.DataFrame(data = ecfp_fingerprints, index = smiles_list)
-    # Remove erroneous data
+
     if len(erroneous_smiles)>0:
         print("The following erroneous SMILES have been found in the data:\n{}.\nThe erroneous SMILES will be removed from the data.".format('\n'.join(map(str, erroneous_smiles))))           
         df_ecfp_fingerprints = df_ecfp_fingerprints.dropna(how='any')    
@@ -115,7 +115,7 @@ st.write('**Insert your SMILES**')
 
 st.write('Type any SMILES used as a reactant in the redox reaction. This model will output the reaction energy.')
 
-## Read SMILES input
+
 SMILES_input = "Oc1cccc(c12)c(O)c(nn2)O\nc1cccc(c12)cc(nn2)O\nOc1c(O)ccc(c12)cc(nn2)O"
 
 SMILES = st.text_area('press ctrl+enter to run model!', SMILES_input, height=20)
@@ -124,7 +124,7 @@ SMILES = list(filter(None, SMILES))
 
 
 
-# Use only top 1000
+
 if len(SMILES)>1000:
     SMILES=SMILES[0:1000]
 	
@@ -132,57 +132,48 @@ ecfc_encoder = get_ecfc(SMILES)
 
 generated_dataset = generate(SMILES)
 
-### transformer for gcn 
+
 filename = 'final_models/transformers.pkl'
 infile = open(filename,'rb')
 transformers = pickle.load(infile)
 infile.close()
 
 
-## model for gcn 
+
 model_dir = 'final_models/tf_chp_initial'
 gcne_model = dc.models.GraphConvModel(n_tasks=1, batch_size=100, mode='regression', dropout=0.25,model_dir= model_dir,random_seed=0)
 gcne_model.restore('final_models/tf_chp_initial/ckpt-94/ckpt-197')
-#print(gcne_model)
 
 
-## predict energy from gcn model 
+
+
 pred_gcne = gcne_model.predict(generated_dataset, transformers)
 
 
-#---------------------------------------------------------------------------------
-##keras model load
+
 from keras.models import model_from_json
 
 keras_final_model = model_from_json(open('./final_models/keras_final_model_architecture.json').read())
 keras_final_model.load_weights('./final_models/keras_final_model_weights.h5')
 
-#keras_final_model = pickle.load(open(r'./final_models/keras_final_model.txt', "rb"))
+
 rf_final_model = pickle.load(open(r'./final_models/rf_final_model.txt', "rb"))
-#xgbm_final_model = pickle.load(open(r'.\final_models\xgbm_final_model.txt', "rb"))
 
 
 
-#predict test data (Keras,RF, GCN)
+
+
 pred_keras = keras_final_model.predict(ecfc_encoder)   
 pred_rf  = rf_final_model.predict(ecfc_encoder)
 
-##reshape (n,)    ----> (n,1)
-
-pred_rf_r = pred_rf.reshape((len(pred_rf),1))
-#pred_xgb = xgbm_final_model.predict(ecfc_encoder)   
 
 
-#calculate consensus
-pred_consensus = (pred_keras + pred_gcne + pred_rf)/3
-# predefined_models.get_errors(test_logS_list,pred_enseble)
+pred_rf_r = pred_rf.reshape((len(pred_rf),1))
 
-#%% Weighted 
 
-#------------------------------------------------------------------------------------------------------------------
 
 
-#------------------------------------------------------------------------------------------------------------------
+pred_consensus = (pred_keras + pred_gcne + pred_rf)/3
 
 
 
@@ -191,23 +182,23 @@ pred_consensus = (pred_keras + pred_gcne + pred_rf)/3
 
 from sklearn.metrics import mean_absolute_error,mean_squared_error,r2_score
 
-## Test 1 Experiments
+
 
 test1_mae = []
 
-test1_mae.append(0.00705) # 0 - GCN
-test1_mae.append(0.00416) # 1 - Keras
-test1_mae.append(0.0035) # 3 - RF
+test1_mae.append(0.00705) 
+test1_mae.append(0.00416) 
+test1_mae.append(0.0035) 
+
 
 
 
-## Test 2 Experiments
 
 test2_mae = []
 
-test2_mae.append(0.00589) # 0 - GCN
-test2_mae.append(0.00483) # 1 - Keras
-test2_mae.append(0.00799) # 3 - RF
+test2_mae.append(0.00589) 
+test2_mae.append(0.00483) 
+test2_mae.append(0.00799) 
 
 
 
@@ -218,9 +209,7 @@ weighted_pred_0_1_3=( np.power(2/(test1_mae[0]+test2_mae[0]),3) * pred_gcne +
 
 
 
-#--------
 
-#### ????  array shape not correct and no difference with pred_consensus
 
 pred_weighted = (pred_gcne + pred_keras + pred_rf_r)/3
 
@@ -230,31 +219,17 @@ pred_weighted = (pred_gcne + pred_keras + pred_rf_r)/3
 
 
 
-#%%
-# results=np.column_stack([pred_mlp,pred_xgb,pred_rf,pred_consensus])
+
 
 df_results = pd.DataFrame(SMILES, columns=['SMILES Reactant'])
 df_results["Predicted Reaction Energy"]= weighted_pred_0_1_3
-#df_results["reaction_energy"]= pred_weighted
-df_results=df_results.round(6)
 
-# df_results.to_csv("results/predicted-"+test_data_name+".csv",index=False)
+df_results=df_results.round(6)
 
 
-# Results DF
 
 st.header('Prediction of Reaction Energy for RFB')
-df_results # Skips the dummy first item
-
-
-
-
-
+df_results 
 
-# download=st.button('Download Results File')
-# if download:
-csv = df_results.to_csv(index=False)
-b64 = base64.b64encode(csv.encode()).decode()  # some strings
-