Update app.py

app.py

@@ -11,18 +11,18 @@ tokenizer = GPT2Tokenizer.from_pretrained("distilgpt2")
 model = GPT2LMHeadModel.from_pretrained("distilgpt2")
 model.to(device)
 
-# Load summarization model (e.g., T5-small)
-summarizer_tokenizer = AutoTokenizer.from_pretrained("t5-small")
-summarizer_model = AutoModelForSeq2SeqLM.from_pretrained("t5-small").to(device)
+# # Load summarization model (e.g., T5-small)
+# summarizer_tokenizer = AutoTokenizer.from_pretrained("t5-small")
+# summarizer_model = AutoModelForSeq2SeqLM.from_pretrained("t5-small").to(device)
 
-def summarize_history(history):
-    input_ids = summarizer_tokenizer.encode(
-        "summarize: " + history,
-        return_tensors="pt"
-    ).to(device)
-    summary_ids = summarizer_model.generate(input_ids, max_length=50, min_length=25, length_penalty=5., num_beams=2)
-    summary = summarizer_tokenizer.decode(summary_ids[0], skip_special_tokens=True)
-    return summary
+# def summarize_history(history):
+#     input_ids = summarizer_tokenizer.encode(
+#         "summarize: " + history,
+#         return_tensors="pt"
+#     ).to(device)
+#     summary_ids = summarizer_model.generate(input_ids, max_length=50, min_length=25, length_penalty=5., num_beams=2)
+#     summary = summarizer_tokenizer.decode(summary_ids[0], skip_special_tokens=True)
+#     return summary
 
 # Set up conversational memory using LangChain's ConversationBufferMemory
 memory = ConversationBufferMemory()
@@ -32,9 +32,9 @@ def chat_with_distilgpt2(input_text):
     # Retrieve conversation history
     conversation_history = memory.load_memory_variables({})['history']
     
-    # Summarize if history exceeds certain length
-    if len(conversation_history.split()) > 200:
-        conversation_history = summarize_history(conversation_history)
+    # # Summarize if history exceeds certain length
+    # if len(conversation_history.split()) > 200:
+    #     conversation_history = summarize_history(conversation_history)
     
     # Combine the (possibly summarized) history with the current user input
     full_input = f"{conversation_history}\nUser: {input_text}\nAssistant:"
@@ -50,9 +50,9 @@ def chat_with_distilgpt2(input_text):
         num_return_sequences=1,
         no_repeat_ngram_size=3,
         repetition_penalty=1.2,
-        temperature=0.9,
-        top_k=20,
-        top_p=0.8,
+        # temperature=0.9,
+        # top_k=20,
+        # top_p=0.8,
         early_stopping=True,
         pad_token_id=tokenizer.eos_token_id,
         eos_token_id=tokenizer.eos_token_id