question generation

app.py

@@ -24,21 +24,21 @@ login(os.environ["HF_TOKEN"])
 dt = datetime.datetime.now()
 print(dt)
 print("loading models")
-# tokenizer = GPT2Tokenizer.from_pretrained('microsoft/DialoGPT-medium')
-# original_model = GPT2LMHeadModel.from_pretrained('microsoft/DialoGPT-medium')
-# untethered_model = GPT2LMHeadModel.from_pretrained('zmbfeng/untethered_20240225_epochs_500')
-# question_generation_tokenizer = AutoTokenizer.from_pretrained("mrm8488/t5-base-finetuned-question-generation-ap")
-# question_generation_model = AutoModelWithLMHead.from_pretrained("mrm8488/t5-base-finetuned-question-generation-ap")
-# paraphrase_tokenizer = AutoTokenizer.from_pretrained("Vamsi/T5_Paraphrase_Paws")
-# paraphrase_model = AutoModelForSeq2SeqLM.from_pretrained("Vamsi/T5_Paraphrase_Paws")
-
-tokenizer = GPT2Tokenizer.from_pretrained('microsoft/DialoGPT-medium',cache_dir="G:\My Drive\Avatar\language_models_windows")
-original_model = GPT2LMHeadModel.from_pretrained('microsoft/DialoGPT-medium',cache_dir="G:\My Drive\Avatar\language_models_windows")
-untethered_model = GPT2LMHeadModel.from_pretrained('zmbfeng/untethered_20240225_epochs_500',cache_dir="G:\My Drive\Avatar\language_models_windows")
-question_generation_tokenizer = AutoTokenizer.from_pretrained("mrm8488/t5-base-finetuned-question-generation-ap",cache_dir="G:\\My Drive\\Avatar\\language_models_windows")
-question_generation_model = AutoModelWithLMHead.from_pretrained("mrm8488/t5-base-finetuned-question-generation-ap",cache_dir="G:\\My Drive\\Avatar\\language_models_windows")
-paraphrase_tokenizer = AutoTokenizer.from_pretrained("Vamsi/T5_Paraphrase_Paws",cache_dir="G:\\My Drive\\Avatar\\language_models_windows")
-paraphrase_model = AutoModelForSeq2SeqLM.from_pretrained("Vamsi/T5_Paraphrase_Paws",cache_dir="G:\\My Drive\\Avatar\\language_models_windows")
+tokenizer = GPT2Tokenizer.from_pretrained('microsoft/DialoGPT-medium')
+original_model = GPT2LMHeadModel.from_pretrained('microsoft/DialoGPT-medium')
+untethered_model = GPT2LMHeadModel.from_pretrained('zmbfeng/untethered_20240225_epochs_500')
+question_generation_tokenizer = AutoTokenizer.from_pretrained("mrm8488/t5-base-finetuned-question-generation-ap")
+question_generation_model = AutoModelWithLMHead.from_pretrained("mrm8488/t5-base-finetuned-question-generation-ap")
+paraphrase_tokenizer = AutoTokenizer.from_pretrained("Vamsi/T5_Paraphrase_Paws")
+paraphrase_model = AutoModelForSeq2SeqLM.from_pretrained("Vamsi/T5_Paraphrase_Paws")
+
+# tokenizer = GPT2Tokenizer.from_pretrained('microsoft/DialoGPT-medium',cache_dir="G:\My Drive\Avatar\language_models_windows")
+# original_model = GPT2LMHeadModel.from_pretrained('microsoft/DialoGPT-medium',cache_dir="G:\My Drive\Avatar\language_models_windows")
+# untethered_model = GPT2LMHeadModel.from_pretrained('zmbfeng/untethered_20240225_epochs_500',cache_dir="G:\My Drive\Avatar\language_models_windows")
+# question_generation_tokenizer = AutoTokenizer.from_pretrained("mrm8488/t5-base-finetuned-question-generation-ap",cache_dir="G:\\My Drive\\Avatar\\language_models_windows")
+# question_generation_model = AutoModelWithLMHead.from_pretrained("mrm8488/t5-base-finetuned-question-generation-ap",cache_dir="G:\\My Drive\\Avatar\\language_models_windows")
+# paraphrase_tokenizer = AutoTokenizer.from_pretrained("Vamsi/T5_Paraphrase_Paws",cache_dir="G:\\My Drive\\Avatar\\language_models_windows")
+# paraphrase_model = AutoModelForSeq2SeqLM.from_pretrained("Vamsi/T5_Paraphrase_Paws",cache_dir="G:\\My Drive\\Avatar\\language_models_windows")
 
 # tokenizer = GPT2Tokenizer.from_pretrained('microsoft/DialoGPT-medium',cache_dir="C:\\Users\\zmbfeng\\Google Drive\\language_models_windows")
 # original_model = GPT2LMHeadModel.from_pretrained('microsoft/DialoGPT-medium',cache_dir="C:\\Users\\zmbfeng\\Google Drive\\Avatar\\language_models_windows")
@@ -49,6 +49,18 @@ paraphrase_model = AutoModelForSeq2SeqLM.from_pretrained("Vamsi/T5_Paraphrase_Pa
 # paraphrase_model = AutoModelForSeq2SeqLM.from_pretrained("Vamsi/T5_Paraphrase_Paws",cache_dir="C:\\Users\\zmbfeng\\Google Drive\\Avatar\\language_models_windows")
 default_temperature=0.01
 default_seed=43
+def create_response_question_generation(input_str, max_length=64):
+    input_text = "answer: %s  context: %s </s>" % (input_str, input_str)
+    print(f"create question input_text={input_text}")
+    features = question_generation_tokenizer([input_text], return_tensors='pt')
+
+    output = question_generation_model.generate(input_ids=features['input_ids'],
+                                                attention_mask=features['attention_mask'],
+                                                max_length=max_length)
+
+    return question_generation_tokenizer.decode(output[0])
+
+
 def create_response(input_str,
                              temperature,
                              seed,
@@ -82,23 +94,14 @@ def create_response(input_str,
 
 common_examples_string="<br/>Sample Inputs:<br/>What is death?<br/>One of the best teachers in all of life turns out to be what?<br/>what is your most meaningful relationship?<br/>What actually gives life meaning?<br/>"
 
-interface_original = gr.Interface(fn=create_response,
-    title="original",
-    description="original language model, no fine tuning"+common_examples_string,
-    #examples=examples,
+interface_original = gr.Interface(fn=create_response_question_generation,
+    title="Question Generation",
+    description="Enter a statment like Paris is the captial of France",
     inputs=[
     gr.Textbox(label="input text here", lines=3),
-    # gr.Number(label="num_beams (integer) explores the specified number of possible outputs and selects the most " +
-    #             "likely ones (specified in num_beams)", value=7),
-        gr.Number(
-            label="temperature (decimal) controls the creativity or randomness of the output. A higher temperature" +
-                  " (e.g., 1.6) results in more diverse and creative output, while a lower temperature (e.g., 0.02)" +
-                  " makes the output more deterministic and focused",
-            value=default_temperature),
         gr.Number(
-            label="seed (integer) random seed, set to -1 to use a random seed everytime",
-                value=default_seed),
-        gr.Textbox(label="model", lines=3, value="original_model",visible=False)
+            label="max length",
+                value=64),
     ],
     outputs="html"
     )