Update README.md

README.md

@@ -11,33 +11,84 @@ probably proofread and complete it, then remove this comment. -->
 
 # t5-base-korean-chit-chat
 
-This model was trained from scratch on an unknown dataset.
-It achieves the following results on the evaluation set:
-
-
-## Model description
-
-More information needed
-
-## Intended uses & limitations
-
-More information needed
-
-## Training and evaluation data
-
-More information needed
-
-## Training procedure
-
-### Training hyperparameters
-
-The following hyperparameters were used during training:
-- optimizer: None
-- training_precision: float32
-
-### Training results
-
-
+This model is a fine-tuning of paust/pko-t5-large model using AIHUB "한국어 SNS". This model infers the next conversation by using the conversation used on social media..
+
+이 모델은 paust/pko-t5-large model을 AIHUB "한국어 SNS"를 이용하여 fine tunning 한 것입니다. 이 모델은 SNS상에서 사용되는 대화를 이용하여 다음 대화를 추론 합니다.
+
+
+## Usage
+```python
+
+from transformers import AutoModelForSeq2SeqLM, DataCollatorForSeq2Seq, Seq2SeqTrainingArguments, Seq2SeqTrainer, MT5ForConditionalGeneration
+from transformers import AutoTokenizer, T5TokenizerFast
+import nltk
+nltk.download('punkt')
+
+
+model_dir = f"lcw99/t5-base-korean-chit-chat"
+
+max_input_length = 1024
+
+text = """
+A: 쇼핑하러 갈까? B: 응 좋아. A: 언제 갈까? B:
+"""
+
+inputs = [text]
+
+inputs = tokenizer(inputs, max_length=max_input_length, truncation=True, return_tensors="pt")
+output = model.generate(**inputs, num_beams=3, do_sample=True, min_length=20, max_length=500, num_return_sequences=3)
+for i in range(3):
+    #print(output[i])
+    print("---", i)
+    decoded_output = tokenizer.decode(output[i], skip_special_tokens=True)
+    predicted_title = nltk.sent_tokenize(decoded_output)
+    #print(decoded_output)
+    print(predicted_title)
+
+import torch
+
+chat_history = []
+# Let's chat for 5 lines
+for step in range(100):
+    print("")
+    user_input = input(">> User: ")
+    chat_history.append("A: " + user_input)
+    while len(chat_history) > 5:
+        chat_history.pop(0)
+    # print(chat_history)
+    hist = ""
+    for chat in chat_history:
+        hist += "\n" + chat
+    hist += "\nB: "
+    # print("====", len(chat_history))
+    # print("===>", hist.replace("\n", " / "))
+    # print("----")
+    # encode the new user input, add the eos_token and return a tensor in Pytorch
+    new_user_input_ids = tokenizer.encode(hist, return_tensors='pt')
+    # print(new_user_input_ids)
+
+    # append the new user input tokens to the chat history
+    #bot_input_ids = torch.cat([chat_history_ids, new_user_input_ids], dim=-1) if step > 0 else new_user_input_ids
+    bot_input_ids = new_user_input_ids
+
+    # generated a response while limiting the total chat history to 1000 tokens, 
+    chat_history_ids = model.generate(
+        bot_input_ids, max_length=200,
+        pad_token_id=tokenizer.eos_token_id,  
+        #no_repeat_ngram_size=3,       
+        do_sample=True, 
+        #top_k=100, 
+        #top_p=0.7,
+        #temperature = 0.1
+    )
+
+    bot_text = tokenizer.decode(chat_history_ids[0], skip_special_tokens=True).replace("#@이름#", "OOO")
+    bot_text = bot_text.replace("\n", " / ")
+    chat_history.append("B: " + bot_text)
+    
+    # pretty print last ouput tokens from bot
+    print("Bot: {}".format(bot_text))    
+```
 
 ### Framework versions