added pos/chunk tab, added vicuna strategy fn

app.py

@@ -79,95 +79,188 @@ def gpt_respond(tab_name, message, chat_history, max_convo_length = 10):
     chat_history.append((message, bot_message))
     return "", chat_history
 
-def vicuna_respond(tab_name, message, chat_history, linguistic_features):
+def vicuna_respond(tab_name, message, chat_history):
     formatted_prompt = f'''Generate the output only for the assistant. Please output any {tab_name} in the following sentence one per line without any additional text: {message}'''
-    print('Vicuna - Prompt + Context:')
-    print(formatted_prompt)
+    # print('Vicuna - Prompt + Context:')
+    # print(formatted_prompt)
     input_ids = vicuna_tokenizer.encode(formatted_prompt, return_tensors="pt")
     output_ids = vicuna_model.generate(input_ids, do_sample=True, max_length=1024, num_beams=5, no_repeat_ngram_size=2)
     bot_message = vicuna_tokenizer.decode(output_ids[0], skip_special_tokens=True)
-    print(bot_message)
+    # print(bot_message)
+    
     # Remove formatted prompt from bot_message
     bot_message = bot_message.replace(formatted_prompt, '')
-    print(bot_message)
+    # print(bot_message)
         
     chat_history.append((formatted_prompt, bot_message))
     time.sleep(2)
     return tab_name, "", chat_history
 
-def llama_respond(tab_name, message, chat_history, linguistic_features):
+def llama_respond(tab_name, message, chat_history):
     formatted_prompt = f'''Generate the output only for the assistant. Please output any {tab_name} in the following sentence one per line without any additional text: {message}'''
-    print('Llama - Prompt + Context:')
-    print(formatted_prompt)
+    # print('Llama - Prompt + Context:')
+    # print(formatted_prompt)
     input_ids = llama_tokenizer.encode(formatted_prompt, return_tensors="pt")
     output_ids = llama_model.generate(input_ids, do_sample=True, max_length=1024, num_beams=5, no_repeat_ngram_size=2)
     bot_message = llama_tokenizer.decode(output_ids[0], skip_special_tokens=True)
+
     # Remove formatted prompt from bot_message
     bot_message = bot_message.replace(formatted_prompt, '')
-    print(bot_message)
+    # print(bot_message)
         
     chat_history.append((formatted_prompt, bot_message))
     time.sleep(2)
     return tab_name, "", chat_history
 
-def interface():
-        gr.Markdown(" Description ")
+def vicuna_strategies_respond(strategy, task_name, task_ling_ent, message, chat_history):
+    formatted_prompt = ""
+    if (task_name == "POS Tagging"):
+        if (strategy == "S1"):
+            formatted_prompt = f'''Generate the output only for the assistant. Please output any {task_ling_ent} in the following sentence one per line without any additional text: {message}'''
+        elif (strategy == "S2"):
+            formatted_prompt = f'''Please POS tag the following sentence using Universal POS tag set without generating any additional text: {message}'''
+        elif (strategy == "S3"):
+            formatted_prompt = f'''Please POS tag the following sentence using Universal POS tag set without generating any additional text: {message}'''
+    elif (task_name == "Chunking"):
+        if (strategy == "S1"):
+            formatted_prompt = f'''Generate the output only for the assistant. Please output any {task_ling_ent} in the following sentence one per line without any additional text: {message}'''
+        elif (strategy == "S2"):
+            formatted_prompt = f'''Please chunk the following sentence in CoNLL 2000 format with BIO tags without outputing any additional text: {message}'''
+        elif (strategy == "S3"):
+            formatted_prompt = f'''Please chunk the following sentence in CoNLL 2000 format with BIO tags without outputing any additional text: {message}'''
+    
+    # print('Vicuna - Prompt + Context:')
+    # print(formatted_prompt)
+    input_ids = vicuna_tokenizer.encode(formatted_prompt, return_tensors="pt")
+    output_ids = vicuna_model.generate(input_ids, do_sample=True, max_length=1024, num_beams=5, no_repeat_ngram_size=2)
+    bot_message = vicuna_tokenizer.decode(output_ids[0], skip_special_tokens=True)
+    # print(bot_message)
+    
+    # Remove formatted prompt from bot_message
+    bot_message = bot_message.replace(formatted_prompt, '')
+    # print(bot_message)
+        
+    chat_history.append((formatted_prompt, bot_message))
+    time.sleep(2)
+    return task_name, "", chat_history
 
-        textbox_prompt = gr.Textbox(show_label=False, placeholder="Write a prompt and press enter")
-        with gr.Row():
-            api_key_input = gr.Textbox(label="Open AI Key", placeholder="Enter your Openai key here", type="password")
-            api_key_btn = gr.Button(value="Submit Key", scale=0)
-        tab_name = gr.Dropdown(["Noun", "Determiner", "Noun phrase", "Verb phrase", "Dependent clause", "T-units"], label="Linguistic Entity")
-        btn = gr.Button(value="Submit")
+def interface():
 
         # prompt = template_single.format(tab_name, textbox_prompt)
 
-        gr.Markdown("Strategy 1 QA-Based Prompting")
-
-        linguistic_features_textbox = gr.Textbox(label="Linguistic Features", disabled=True)
-
-        with gr.Row():
-            vicuna_S1_chatbot = gr.Chatbot(label="vicuna-7b")
-            llama_S1_chatbot = gr.Chatbot(label="llama-7b")
-            gpt_S1_chatbot = gr.Chatbot(label="gpt-3.5")
-        clear = gr.ClearButton(components=[textbox_prompt, api_key_input, vicuna_S1_chatbot, llama_S1_chatbot, gpt_S1_chatbot])
-        # gr.Markdown("Strategy 2 Instruction-Based Prompting")
-        # with gr.Row():
-        #     vicuna_S2_chatbot = gr.Chatbot(label="vicuna-7b")
-        #     llama_S2_chatbot = gr.Chatbot(label="llama-7b")
-        #     gpt_S2_chatbot = gr.Chatbot(label="gpt-3.5")
-        # clear = gr.ClearButton(components=[textbox_prompt, vicuna_S2_chatbot])
-        # gr.Markdown("Strategy 3 Structured Prompting")
-        # with gr.Row():
-        #     vicuna_S3_chatbot = gr.Chatbot(label="vicuna-7b")
-        #     llama_S3_chatbot = gr.Chatbot(label="llama-7b")
-        #     gpt_S3_chatbot = gr.Chatbot(label="gpt-3.5")
-        # clear = gr.ClearButton(components=[textbox_prompt, vicuna_S3_chatbot])
-
-        #textbox_prompt.submit(vicuna_respond, inputs=[textbox_prompt, vicuna_S1_chatbot], outputs=[textbox_prompt, vicuna_S1_chatbot])
+        with gr.Tab("Linguistic Entities"):
+            gr.Markdown(" Description Here ")
+
+            # Inputs
+            ling_ents_prompt = gr.Textbox(show_label=False, placeholder="Write a prompt and press enter")
+            with gr.Row():
+                # Will activate after getting API key
+                ling_ents_apikey_input = gr.Textbox(label="Open AI Key", placeholder="Enter your Openai key here", type="password")
+                ling_ents_apikey_btn = gr.Button(value="Submit Key", scale=0)
+            linguistic_entities = gr.Dropdown(["Noun", "Determiner", "Noun phrase", "Verb phrase", "Dependent clause", "T-units"], label="Linguistic Entity")
+            ling_ents_btn = gr.Button(value="Submit")
+
+            # Outputs
+            gr.Markdown("Strategy 1 QA-Based Prompting")
+
+            linguistic_features_textbox = gr.Textbox(label="Linguistic Features", disabled=True)
+
+            with gr.Row():
+                vicuna_ling_ents_chatbot = gr.Chatbot(label="vicuna-7b")
+                llama_ling_ents_chatbot = gr.Chatbot(label="llama-7b")
+                gpt_ling_ents_chatbot = gr.Chatbot(label="gpt-3.5")
+            clear = gr.ClearButton(components=[ling_ents_prompt, ling_ents_apikey_input, vicuna_ling_ents_chatbot, llama_ling_ents_chatbot, gpt_ling_ents_chatbot])
+
+            # Event Handler for Vicuna Chatbot
+            ling_ents_btn.click(vicuna_respond, inputs=[linguistic_entities, ling_ents_prompt, vicuna_ling_ents_chatbot], 
+                      outputs=[linguistic_entities, ling_ents_prompt, vicuna_ling_ents_chatbot])
+            
+            # Event Handler for LLaMA Chatbot
+            ling_ents_btn.click(llama_respond, inputs=[linguistic_entities, ling_ents_prompt, llama_ling_ents_chatbot], 
+                      outputs=[linguistic_entities, ling_ents_prompt, llama_ling_ents_chatbot])
+            
+            # Event Handler for GPT 3.5 Chatbot, user must submit api key before submitting the prompt
+            # Will activate after getting API key
+            # ling_ents_apikey_btn.click(update_api_key, inputs=ling_ents_apikey_input)
+            # ling_ents_btn.click(gpt_respond, inputs=[linguistic_entities, ling_ents_prompt, gpt_ling_ents_chatbot], 
+            #                     outputs=[linguistic_entities, ling_ents_prompt, gpt_ling_ents_chatbot])
+
+        with gr.Tab("POS/Chunking"):
+            gr.Markdown(" Description Here ")
+
+            # Inputs
+            task_prompt = gr.Textbox(show_label=False, placeholder="Write a prompt and press enter")
+            with gr.Row():
+                task_apikey_input = gr.Textbox(label="Open AI Key", placeholder="Enter your Openai key here", type="password")
+                task_apikey_btn = gr.Button(value="Submit Key", scale=0)
+            task = gr.Dropdown(["POS Tagging", "Chunking"], label="Task")
+            task_linguistic_entities = gr.Dropdown(["Noun", "Determiner", "Noun phrase", "Verb phrase", "Dependent clause", "T-units"], label="Linguistic Entity For Strategy 1")
+            task_btn = gr.Button(value="Submit")
+
+            # Outputs
+            gr.Markdown("Strategy 1 QA-Based Prompting")
+            with gr.Row():
+                vicuna_S1_chatbot = gr.Chatbot(label="vicuna-7b")
+                llama_S1_chatbot = gr.Chatbot(label="llama-7b")
+                gpt_S1_chatbot = gr.Chatbot(label="gpt-3.5")
+            gr.Markdown("Strategy 2 Instruction-Based Prompting")
+            with gr.Row():
+                vicuna_S2_chatbot = gr.Chatbot(label="vicuna-7b")
+                llama_S2_chatbot = gr.Chatbot(label="llama-7b")
+                gpt_S2_chatbot = gr.Chatbot(label="gpt-3.5")
+            gr.Markdown("Strategy 3 Structured Prompting")
+            with gr.Row():
+                vicuna_S3_chatbot = gr.Chatbot(label="vicuna-7b")
+                llama_S3_chatbot = gr.Chatbot(label="llama-7b")
+                gpt_S3_chatbot = gr.Chatbot(label="gpt-3.5")
+            clear_all = gr.ClearButton(components=[task_prompt, task_apikey_input,
+                                               vicuna_S1_chatbot, llama_S1_chatbot, gpt_S1_chatbot, 
+                                               vicuna_S2_chatbot, llama_S2_chatbot, gpt_S2_chatbot,
+                                               vicuna_S3_chatbot, llama_S3_chatbot, gpt_S3_chatbot])
+            
+            # vicuna_strategies_respond(strategy, task_name, task_ling_ent, message, chat_history):
+            # Event Handlers for Vicuna Chatbot POS/Chunk
+            task_btn.click(vicuna_strategies_respond, inputs=["S1", task, task_linguistic_entities, task_prompt, vicuna_S1_chatbot], 
+                      outputs=[task, task_prompt, vicuna_S1_chatbot])
+            task_btn.click(vicuna_strategies_respond, inputs=["S2", task, task_linguistic_entities, task_prompt, vicuna_S2_chatbot], 
+                      outputs=[task, task_prompt, vicuna_S2_chatbot])
+            task_btn.click(vicuna_strategies_respond, inputs=["S3", task, task_linguistic_entities, task_prompt, vicuna_S3_chatbot], 
+                      outputs=[task, task_prompt, vicuna_S3_chatbot])
+            
+            # Event Handler for LLaMA Chatbot POS/Chunk
+            task_btn.click(llama_respond, inputs=[task, task_prompt, llama_S1_chatbot], 
+                      outputs=[task, task_prompt, llama_S1_chatbot])
+            task_btn.click(llama_respond, inputs=[task, task_prompt, llama_S2_chatbot], 
+                      outputs=[task, task_prompt, llama_S2_chatbot])
+            task_btn.click(llama_respond, inputs=[task, task_prompt, llama_S3_chatbot], 
+                      outputs=[task, task_prompt, llama_S3_chatbot])
+            
+            # Event Handler for GPT 3.5 Chatbot, user must submit api key before submitting the prompt
+            # Will activate after getting API key
+            # task_apikey_btn.click(update_api_key, inputs=ling_ents_apikey_input)
+            # task_btn.click(gpt_respond, inputs=[task, task_prompt, gpt_S1_chatbot], 
+            #                     outputs=[task, task_prompt, gpt_S1_chatbot])
+            # task_btn.click(gpt_respond, inputs=[task, task_prompt, gpt_S2_chatbot], 
+            #                     outputs=[task, task_prompt, gpt_S2_chatbot])
+            # task_btn.click(gpt_respond, inputs=[task, task_prompt, gpt_S3_chatbot], 
+            #                     outputs=[task, task_prompt, gpt_S3_chatbot])
+            
+
+        # textbox_prompt.submit(vicuna_respond, inputs=[textbox_prompt, vicuna_S1_chatbot], outputs=[textbox_prompt, vicuna_S1_chatbot])
         # textbox_prompt.submit(respond, inputs=[textbox_prompt, vicuna_S2_chatbot], outputs=[textbox_prompt, vicuna_S2_chatbot])
         # textbox_prompt.submit(respond, inputs=[textbox_prompt, vicuna_S3_chatbot], outputs=[textbox_prompt, vicuna_S3_chatbot])
 
         #textbox_prompt.submit(llama_respond, inputs=[textbox_prompt, llama_S1_chatbot], outputs=[textbox_prompt, llama_S1_chatbot])
 
-        btn.click(lambda _, 
-                  message=textbox_prompt: linguistic_features_textbox.update(linguistic_features(textbox_prompt.value)),
-                  inputs=[textbox_prompt], 
-                  outputs=[linguistic_features_textbox])
-
-        
-        btn.click(vicuna_respond, inputs=[tab_name, textbox_prompt, vicuna_S1_chatbot], 
-                  outputs=[tab_name, textbox_prompt, vicuna_S1_chatbot])
-        
-        btn.click(llama_respond, inputs=[tab_name, textbox_prompt, llama_S1_chatbot], 
-                  outputs=[tab_name, textbox_prompt, llama_S1_chatbot])
-
-        #api_key_btn.click(update_api_key, inputs=api_key_input)
-        #btn.click(gpt_respond, inputs=[tab_name, textbox_prompt, gpt_S1_chatbot], outputs=[tab_name, textbox_prompt, gpt_S1_chatbot])
+        # btn.click(lambda _, 
+        #           message=textbox_prompt: linguistic_features_textbox.update(linguistic_features(textbox_prompt.value)),
+        #           inputs=[textbox_prompt], 
+        #           outputs=[linguistic_features_textbox])
 
 with gr.Blocks() as demo:
     gr.Markdown("# LLM Evaluator With Linguistic Scrutiny")
 
+    # load interface
     interface()
     
 demo.launch()