Update app.py

app.py

@@ -6,6 +6,7 @@ from transformers import (
     AutoTokenizer
 )
 
+M0 = "anshoomehra/question-generation-auto-t5-v1-base-s"
 M1 = "anshoomehra/question-generation-auto-t5-v1-base-s-q"
 M2 = "anshoomehra/question-generation-auto-t5-v1-base-s-q-c" 
 
@@ -14,6 +15,9 @@ M5 = "anshoomehra/question-generation-auto-hints-t5-v1-base-s-q-c"
 
 device = ['cuda' if torch.cuda.is_available() else 'cpu'][0]
 
+_m0 = AutoModelForSeq2SeqLM.from_pretrained(M0).to(device)
+_tk0 = AutoTokenizer.from_pretrained(M0, cache_dir="./cache")
+
 _m1 = AutoModelForSeq2SeqLM.from_pretrained(M1).to(device)
 _tk1 = AutoTokenizer.from_pretrained(M1, cache_dir="./cache")
 
@@ -48,6 +52,7 @@ def _formatQs(questions):
     
 def _generate(mode, context, hint=None, minLength=50, maxLength=500, lengthPenalty=2.0, earlyStopping=True, numReturnSequences=1, numBeams=2, noRepeatNGramSize=0, doSample=False, topK=0, topP=0, temperature=0):
           
+    predictionM0 = None
     predictionM1 = None
     predictionM2 = None
     predictionM4 = None
@@ -55,7 +60,23 @@ def _generate(mode, context, hint=None, minLength=50, maxLength=500, lengthPenal
     
     if mode == 'Auto':
         _inputText = "question_context: " + context
-        
+
+        _encoding = _tk0.encode(_inputText, return_tensors='pt', truncation=True, padding='max_length').to(device) # max_length=1024 
+        _outputEncoded = _m0.generate(_encoding, 
+                                   min_length=minLength, 
+                                   max_length=maxLength,
+                                   length_penalty=lengthPenalty,
+                                   early_stopping=earlyStopping,
+                                   num_return_sequences=numReturnSequences,
+                                   num_beams=numBeams,
+                                   no_repeat_ngram_size=noRepeatNGramSize,
+                                   do_sample=doSample,
+                                   top_k=topK,
+                                   top_p=topP,
+                                   temperature=temperature
+                               )
+        predictionM0 = [_tk0.decode(id, clean_up_tokenization_spaces=False, skip_special_tokens=True) for id in _outputEncoded]
+              
         _encoding = _tk1.encode(_inputText, return_tensors='pt', truncation=True, padding='max_length').to(device) # max_length=1024 
         _outputEncoded = _m1.generate(_encoding, 
                                    min_length=minLength, 
@@ -119,7 +140,6 @@ def _generate(mode, context, hint=None, minLength=50, maxLength=500, lengthPenal
                                    temperature=temperature
                                )
         predictionM5 = [_tk5.decode(id, clean_up_tokenization_spaces=False, skip_special_tokens=True) for id in _outputEncoded]
- 
     elif mode == 'Hints':
         _inputText = "question_hint: " + hint + "</s>question_context: " + context
 
@@ -155,12 +175,13 @@ def _generate(mode, context, hint=None, minLength=50, maxLength=500, lengthPenal
                                )
         predictionM5 = [_tk5.decode(id, clean_up_tokenization_spaces=False, skip_special_tokens=True) for id in _outputEncoded]
   
+    predictionM0 = _formatQs(predictionM0)
     predictionM1 = _formatQs(predictionM1)
     predictionM2 = _formatQs(predictionM2)
     predictionM4 = _formatQs(predictionM4)
     predictionM5 = _formatQs(predictionM5)
-    
-    return predictionM1, predictionM2, predictionM4, predictionM5
+
+    return predictionM5, predictionM4, predictionM2, predictionM1, predictionM0
     
 with gr.Blocks() as demo:
     
@@ -199,12 +220,13 @@ with gr.Blocks() as demo:
         with gr.Row(variant='compact'):
             _predictionM2 = gr.Textbox(label="Predicted Questions - question-generation-auto-t5-v1-base-s-q-c [No Hints]")
             _predictionM1 = gr.Textbox(label="Predicted Questions - question-generation-auto-t5-v1-base-s-q [No Hints]")
+            _predictionM0 = gr.Textbox(label="Predicted Questions - question-generation-auto-t5-v1-base-s-q [No Hints]")
 
     with gr.Row():       
         gen_btn = gr.Button("Generate Questions")
         gen_btn.click(fn=_generate,
                       inputs=[mode, context, hint, minLength, maxLength, lengthPenalty, earlyStopping, numReturnSequences, numBeams, noRepeatNGramSize, doSample, topK, topP, temperature],
-                      outputs=[_predictionM1, _predictionM2, _predictionM4, _predictionM5]
+                      outputs=[_predictionM5, _predictionM4, _predictionM2, _predictionM1, _predictionM0]
                       )
 
 demo.launch(show_error=True)