Update app.py

app.py

@@ -350,17 +350,44 @@ windows_size = 10
 with open('model_google.pkl', 'rb') as f:
     Mode = pickle.load(f)
 
-def Test_model(text, Model):
-  word_list = text_to_wordlist(text)
-  sequences = tokenizer.texts_to_sequences([word_list])
-  sequences_input = list(itertools.chain(*sequences))
-  sequences_input =  pad_sequences([sequences_input], value=0, padding="post", maxlen=windows_size).tolist()
-  input_a = np.asarray(sequences_input)
-  pred = Model.predict(input_a, batch_size=None, verbose=0, steps=None)
+#def Test_model(text, Model):
+ # word_list = text_to_wordlist(text)
+ # sequences = tokenizer.texts_to_sequences([word_list])
+ # sequences_input = list(itertools.chain(*sequences))
+ # sequences_input =  pad_sequences([sequences_input], value=0, padding="post", maxlen=windows_size).tolist()
+ # input_a = np.asarray(sequences_input)
+ # pred = Model.predict(input_a, batch_size=None, verbose=0, steps=None)
   #print(pred)
-  predicted_class = np.argmax(pred)
+  #predicted_class = np.argmax(pred)
   #print(labels[predicted_class])
 
+def Test_model(text, model, window_size=10):
+    #print(text)
+    word_list = text_to_wordlist(text)
+    #print(word_list)
+    sequences = tokenizer.texts_to_sequences([word_list])
+    sequences_input = list(itertools.chain(*sequences))
+    
+    if len(sequences_input) <= window_size:
+        sequences_input = pad_sequences([sequences_input], value=0, padding="post", maxlen=window_size).tolist()
+        #print(sequences_input)
+        input_a = np.asarray(sequences_input)
+        pred = Modell.predict(input_a, batch_size=None, verbose=0, steps=None)
+        #print(pred)
+        predicted_class = np.argmax(pred)
+        #print(labels[predicted_class])
+    else:
+        predictions = []
+        for i in range(len(sequences_input) - window_size + 1):
+            window_input = sequences_input[i : i + window_size]
+            #print(window_input)
+            input_a = np.asarray([window_input])
+            pred = Modell.predict(input_a, batch_size=None, verbose=0, steps=None)
+            #print(pred)
+            predictions.append(pred)
+        accumulated_pred = np.sum(predictions, axis=0)
+        predicted_class = np.argmax(np.sum(accumulated_pred, axis=0))
+        #print(labels[predicted_class])
 
 
 import gradio as gr
@@ -375,11 +402,27 @@ def predict(text):
   word_list = text_to_wordlist(text)
   sequences = tokenizer.texts_to_sequences([word_list])
   sequences_input = list(itertools.chain(*sequences))
-  sequences_input =  pad_sequences([sequences_input], value=0, padding="post", maxlen=windows_size).tolist()
-  input_a = np.asarray(sequences_input)
-  pred = Modell.predict(input_a, batch_size=None, verbose=0, steps=None)
-  
-  predicted_class = np.argmax(pred)
+    if len(sequences_input) <= window_size:
+        sequences_input = pad_sequences([sequences_input], value=0, padding="post", maxlen=window_size).tolist()
+        #print(sequences_input)
+        input_a = np.asarray(sequences_input)
+        pred = Modell.predict(input_a, batch_size=None, verbose=0, steps=None)
+        #print(pred)
+        predicted_class = np.argmax(pred)
+        #print(labels[predicted_class])
+    else:
+        predictions = []
+        for i in range(len(sequences_input) - window_size + 1):
+            window_input = sequences_input[i : i + window_size]
+            #print(window_input)
+            input_a = np.asarray([window_input])
+            pred = Modell.predict(input_a, batch_size=None, verbose=0, steps=None)
+            #print(pred)
+            predictions.append(pred)
+        accumulated_pred = np.sum(predictions, axis=0)
+        predicted_class = np.argmax(np.sum(accumulated_pred, axis=0))
+        #print(labels[predicted_class])
+
   return labels[predicted_class]
 input_text = gr.inputs.Textbox(label="Enter a sentence")
 output_text = gr.outputs.Textbox(label="Predicted label")