Update app.py

Microsoft DialoGPT model

app.py

@@ -371,23 +371,81 @@
 #-----------------------------------------------------------------------------------
 # 16. Text-to-Text Generation using the T5 model - Task 7 check whether a statement deduced from a text is correct or not.
 
-from transformers import T5ForConditionalGeneration, T5Tokenizer
+# from transformers import T5ForConditionalGeneration, T5Tokenizer
+# import gradio as grad
+
+# text2text_tkn= T5Tokenizer.from_pretrained("t5-small")
+# mdl = T5ForConditionalGeneration.from_pretrained("t5-small")
+
+# def text2text_deductible(sentence1,sentence2):
+#     inp1 = "rte sentence1: "+sentence1
+#     inp2 = "sentence2: "+sentence2
+#     combined_inp=inp1+" "+inp2
+#     enc = text2text_tkn(combined_inp, return_tensors="pt")
+#     tokens = mdl.generate(**enc)
+#     response=text2text_tkn.batch_decode(tokens)
+#     return response
+
+# sent1=grad.Textbox(lines=1, label="Sentence1", placeholder="Text in English")
+# sent2=grad.Textbox(lines=1, label="Sentence2", placeholder="Text in English")
+# out=grad.Textbox(lines=1, label="Whether sentence2 is deductible from sentence1")
+
+# grad.Interface(text2text_deductible, inputs=[sent1,sent2], outputs=out).launch()
+
+#-----------------------------------------------------------------------------------
+# 17. Chatbot/Dialog Bot: a simple bot named Alicia that is based on the Microsoft DialoGPT model .
+
+from transformers import AutoModelForCausalLM, AutoTokenizer,BlenderbotForConditionalGeneration
+import torch
+
+chat_tkn = AutoTokenizer.from_pretrained("microsoft/DialoGPT-medium")
+mdl = AutoModelForCausalLM.from_pretrained("microsoft/DialoGPT-medium")
+
+#chat_tkn = AutoTokenizer.from_pretrained("facebook/blenderbot-400M-distill")
+#mdl = BlenderbotForConditionalGeneration.from_pretrained("facebook/blenderbot-400M-distill")
+
+def converse(user_input, chat_history=[]):
+    user_input_ids = chat_tkn(user_input + chat_tkn.eos_token, return_tensors='pt').input_ids
+    # keep history in the tensor
+    bot_input_ids = torch.cat([torch.LongTensor(chat_history), user_input_ids], dim=-1)
+    # get response
+    chat_history = mdl.generate(bot_input_ids, max_length=1000, pad_token_id=chat_tkn.eos_token_id).tolist()
+    print (chat_history)
+    response = chat_tkn.decode(chat_history[0]).split("<|endoftext|>")
+    print("starting to print response")
+    print(response)
+    # html for display
+    html = "<div class='mybot'>"
+    for x, mesg in enumerate(response):
+        if x%2!=0 :
+           mesg="Alicia:"+mesg
+           clazz="alicia"
+        else :
+           clazz="user"
+        print("value of x")
+        print(x)
+        print("message")
+        print (mesg)
+        html += "<div class='mesg {}'> {}</div>".format(clazz, mesg)
+    html += "</div>"
+    print(html)
+    return html, chat_history
+
 import gradio as grad
 
-text2text_tkn= T5Tokenizer.from_pretrained("t5-small")
-mdl = T5ForConditionalGeneration.from_pretrained("t5-small")
+css = """
+.mychat {display:flex;flex-direction:column}
+.mesg {padding:5px;margin-bottom:5px;border-radius:5px;width:75%}
+.mesg.user {background-color:lightblue;color:white}
+.mesg.alicia {background-color:orange;color:white,align-self:self-end}
+.footer {display:none !important}
+"""
 
-def text2text_deductible(sentence1,sentence2):
-    inp1 = "rte sentence1: "+sentence1
-    inp2 = "sentence2: "+sentence2
-    combined_inp=inp1+" "+inp2
-    enc = text2text_tkn(combined_inp, return_tensors="pt")
-    tokens = mdl.generate(**enc)
-    response=text2text_tkn.batch_decode(tokens)
-    return response
+text=grad.inputs.Textbox(placeholder="Lets chat")
 
-sent1=grad.Textbox(lines=1, label="Sentence1", placeholder="Text in English")
-sent2=grad.Textbox(lines=1, label="Sentence2", placeholder="Text in English")
-out=grad.Textbox(lines=1, label="Whether sentence2 is deductible from sentence1")
+grad.Interface(fn=converse,
+             theme="default",
+             inputs=[text, "state"],
+             outputs=["html", "state"],
+             css=css).launch()
 
-grad.Interface(text2text_deductible, inputs=[sent1,sent2], outputs=out).launch()