swpa out chat logic

app.py

@@ -7,17 +7,25 @@ import torch
 import gradio as gr
 
 
-def chat(message, history):
-    history = history if history is not None else []
-    new_user_input_ids = tokenizer.encode(message+tokenizer.eos_token, return_tensors='pt')
+# def chat(message, history):
+#     history = history if history is not None else []
+#     new_user_input_ids = tokenizer.encode(message+tokenizer.eos_token, return_tensors='pt')
+#     bot_input_ids = torch.cat([torch.LongTensor(history), new_user_input_ids], dim=-1)
+#     history = model.generate(bot_input_ids, max_length=500, pad_token_id=tokenizer.eos_token_id).tolist()
+#     # response = tokenizer.decode(history[0]).replace("<|endoftext|>", "\n")
+#     # pretty print last ouput tokens from bot
+#     response =  tokenizer.decode(bot_input_ids.shape[-1][0], skip_special_tokens=True)
+#     print("The response is ", [response])
+#     # history.append((message, response, new_user_input_ids, chat_history_ids))
+#     return response, history, feedback(message)
+
+
+def chat(message, history=[]):
+    new_user_input_ids = tokenizer.encode(input + tokenizer.eos_token, return_tensors='pt')
     bot_input_ids = torch.cat([torch.LongTensor(history), new_user_input_ids], dim=-1)
     history = model.generate(bot_input_ids, max_length=500, pad_token_id=tokenizer.eos_token_id).tolist()
-    response = tokenizer.decode(history[0]).replace("<|endoftext|>", "\n")
-    # pretty print last ouput tokens from bot
-    # response =  tokenizer.decode(chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)
-    print("The response is ", [response])
-    # history.append((message, response, new_user_input_ids, chat_history_ids))
-    return response, feedback(message)
+    response = tokenizer.decode(history[0]).replace("<|endoftext|>", "")
+    return response, history
 
 
 def feedback(text):