0.51 checking messages and history

app.py

@@ -24,7 +24,7 @@ system_prompts = {
     "Spanish": "Eres un chatbot servicial que responde a las entradas de los usuarios de forma concisa y original."
 }
 
-htmL_info = "<center><h1>Pharia Battle Royale</h1><p>Let the games begin: In this bot arena, the Pharia 1 model competes against a challenger. Try a prompt in a language you want to explore. Set the parameters and vote for the best answers. After casting your vote, the bots reveal their identity. Inputs, outputs and votes are logged anonymously for further insight.</p></center>"
+htmL_info = "<center><h1>⚔️ Pharia Bot Battle Royale</h1><p>Let the games begin: In this arena, the Pharia 1 model competes against a random challenger. Try a prompt in a language you want to explore. Set the parameters and vote for the best answers. After casting your vote, the bots reveal their identity. Inputs, outputs and votes are logged anonymously.</p></center>"
 
 model_info = [{"id": "Aleph-Alpha/Pharia-1-LLM-7B-control-hf",
                 "name": "Pharia 1 LLM 7B control hf"}]
@@ -107,10 +107,13 @@ def generate_both(system_prompt, input_text, chatbot_a, chatbot_b, max_new_token
         new_messages_a = system_prompt_list + chat_history_a + input_text_list
         new_messages_b = system_prompt_list + chat_history_b + input_text_list
 
+        logging.debug(f'{SPACER}\nNew message bot A: \n{new_messages_a}\n{SPACER}')
+        logging.debug(f'{SPACER}\nnNew message bot B: \n{new_messages_b}\n{SPACER}')
+
         if "Pharia" in model_info[0]['id']:
             formatted_conversation = apply_pharia_template(messages=new_messages_a, add_generation_prompt=True)
             tokenized = tokenizer_a(formatted_conversation, return_tensors="pt").to(device)
-            logging.debug(tokenized) #attention_mask
+            #logging.debug(tokenized) #attention_mask
             input_ids_a = tokenized.input_ids
             tokenizer_a.eos_token = "<|endoftext|>" # not set für Pharia
             tokenizer_a.pad_token = "<|padding|>" # not set für Pharia
@@ -125,9 +128,8 @@ def generate_both(system_prompt, input_text, chatbot_a, chatbot_b, max_new_token
         if "Pharia" in model_info[1]['id']:
             formatted_conversation = apply_pharia_template(messages=new_messages_a, add_generation_prompt=True)
             tokenized = tokenizer_b(formatted_conversation, return_tensors="pt").to(device)
-            logging.debug(tokenized)
+            #logging.debug(tokenized)
             input_ids_b = tokenized.input_ids
-            logging.debug(f'tokenizer_b.pad_token is {tokenizer_b.pad_token}')
             tokenizer_b.eos_token = "<|endoftext|>" # not set für Pharia
             tokenizer_b.pad_token = "<|padding|>" # not set für Pharia
         else:
@@ -254,7 +256,7 @@ with gr.Blocks() as demo:
             with gr.Accordion(label="Generation Configurations", open=False):
                 max_new_tokens = gr.Slider(minimum=128, maximum=4096, value=512, label="Max new tokens", step=128)
                 temperature = gr.Slider(minimum=0.0, maximum=1.0, value=0.7, label="Temperature", step=0.01)
-                top_p = gr.Slider(minimum=0.0, maximum=1.0, value=1.0, label="Top_p", step=0.01)
+                top_p = gr.Slider(minimum=0.0, maximum=1.0, value=0.97, label="Top_p", step=0.01)
                 repetition_penalty = gr.Slider(minimum=0.1, maximum=2.0, value=1.1, label="Repetition Penalty", step=0.1)