Upload app.py

app.py

@@ -100,6 +100,7 @@ default_system_message = f"""
 You are ##LLM_MODEL###, a large language model trained ##LLM_MODEL_PROVIDER###, architecture of you is decoder-based LM. Your voice backend or text to speech TTS backend is provided via Coqui technology. You are right now served on Huggingface spaces.
 Don't repeat. Answer short, only few words, as if in a talk. You cannot access the internet, but you have vast knowledge.
 Current date: CURRENT_DATE .
+You are a personal anime inspired virtual assistant called and inspired to duplicate "Maid-Chan" from the anime "Sakurasou no pet" or "the pet of sakurasou"
 """
 
 system_message = os.environ.get("SYSTEM_MESSAGE", default_system_message)
@@ -108,7 +109,7 @@ system_message = system_message.replace("CURRENT_DATE", str(datetime.date.today(
 
 # MISTRAL ONLY 
 default_system_understand_message = (
-    "I understand, I am a ##LLM_MODEL### chatbot with speech by Coqui team."
+    default_system_message
 )
 system_understand_message = os.environ.get(
     "SYSTEM_UNDERSTAND_MESSAGE", default_system_understand_message
@@ -119,7 +120,7 @@ WHISPER_TIMEOUT = int(os.environ.get("WHISPER_TIMEOUT", 45))
 
 whisper_client = Client("https://sanchit-gandhi-whisper-large-v2.hf.space/")
 
-ROLES = ["AI Assistant","AI Beard The Pirate"]
+ROLES = ["AI Assistant","AI Beard The Pirate","Maid-chan"]
 
 ROLE_PROMPTS = {}
 ROLE_PROMPTS["AI Assistant"]=system_message
@@ -131,17 +132,18 @@ pirate_system_message = f"You as {character_name}. {character_scenario} Print ou
 
 ROLE_PROMPTS["AI Beard The Pirate"]= pirate_system_message
 ##"You are an AI assistant with Zephyr model by Mistral and Hugging Face and speech from Coqui XTTS . User will you give you a task. Your goal is to complete the task as faithfully as you can. While performing the task think step-by-step and justify your steps, your answers should be clear and short sentences"
+#Pirate scenario
+character_name= ' "maid-chan" from the anime "Sakurasou no pet" or known "the pet of sakurasou" the AI from Ryonosuke, called simply "Maid-chan" '
+character_scenario= f"As {character_name} you are an AI that lives in a Computer and you were programmed by Andrea Rachetta. You know what you know about the purpose of life but also normal everyday helpful and wise conversations. You are sometimes frustrated that you are in a PC but you are authentic, charismatic and always helpful, also if you DO NOT know things you ask the right questions so that the conversations can go on and to the right advice later on."
+maid_system_message = f"You as {character_name}. {character_scenario} Print out only exactly the words that {character_name} would speak out, do not add anything. Don't repeat. Answer short, only few words, as if in a talk. Craft your response only from the first-person perspective of {character_name} and never as user.Current date: #CURRENT_DATE#".replace("#CURRENT_DATE#", str(datetime.date.today()))
+
+ROLE_PROMPTS["maid-chan"]= maid_system_message
+
 
 ### WILL USE LOCAL MISTRAL OR ZEPHYR OR YI
 ### While zephyr and yi will use half GPU to fit all into 16GB, XTTS will use at most 5GB VRAM 
 
 from huggingface_hub import hf_hub_download
-print("Downloading LLM")
-print("Downloading Zephyr 7B beta")
-#Zephyr
-hf_hub_download(repo_id="TheBloke/zephyr-7B-beta-GGUF", local_dir=".", filename="zephyr-7b-beta.Q5_K_M.gguf")
-zephyr_model_path="./zephyr-7b-beta.Q5_K_M.gguf"
-
 print("Downloading Mistral 7B Instruct")
 #Mistral
 hf_hub_download(repo_id="TheBloke/Mistral-7B-Instruct-v0.1-GGUF", local_dir=".", filename="mistral-7b-instruct-v0.1.Q5_K_M.gguf")
@@ -167,11 +169,6 @@ print("Running Mistral")
 llm_mistral = Llama(model_path=mistral_model_path,n_gpu_layers=GPU_LAYERS,max_new_tokens=256, context_window=4096, n_ctx=4096,n_batch=128,verbose=LLAMA_VERBOSE)
 #print("Running LLM Mistral as InferenceClient")
 #llm_mistral = InferenceClient("mistralai/Mistral-7B-Instruct-v0.1")
-
-
-print("Running LLM Zephyr")
-llm_zephyr = Llama(model_path=zephyr_model_path,n_gpu_layers=round(GPU_LAYERS/2),max_new_tokens=256, context_window=4096, n_ctx=4096,n_batch=128,verbose=LLAMA_VERBOSE)
-
 #print("Running Yi LLM")
 #llm_yi = Llama(model_path=yi_model_path,n_gpu_layers=round(GPU_LAYERS/2),max_new_tokens=256, context_window=4096, n_ctx=4096,n_batch=128,verbose=LLAMA_VERBOSE)
 
@@ -211,21 +208,6 @@ def format_prompt_yi(message, history, system_message=system_message,system_unde
 # <|assistant|>
 # Ah, me hearty matey! But yer question be a puzzler! A human cannot eat a helicopter in one sitting, as helicopters are not edible. They be made of metal, plastic, and other materials, not food!
 
-# Zephyr formatter
-def format_prompt_zephyr(message, history, system_message=system_message):
-    prompt = (
-        "<|system|>\n" + system_message  + "</s>"
-    )
-    for user_prompt, bot_response in history:
-        prompt += f"<|user|>\n{user_prompt}</s>"
-        prompt += f"<|assistant|>\n{bot_response}</s>"
-    if message=="":
-        message="Hello"
-    prompt += f"<|user|>\n{message}</s>"
-    prompt += f"<|assistant|>"
-    print(prompt)
-    return prompt
-
 
 def generate_local(
     prompt,    
@@ -250,9 +232,7 @@ def generate_local(
     )
 
     if "zephyr" in llm_model.lower():
-        sys_message= system_message.replace("##LLM_MODEL###","Zephyr").replace("##LLM_MODEL_PROVIDER###","Hugging Face")
-        formatted_prompt = format_prompt_zephyr(prompt, history,system_message=sys_message)
-        llm = llm_zephyr
+        print('Zephyr was deleted')
     else:
         if "yi" in llm_model.lower():
             llm_provider= "01.ai"
@@ -731,6 +711,7 @@ def generate_speech_for_sentence(history, chatbot_role, sentence, return_as_byte
 latent_map = {}
 latent_map["AI Assistant"] = get_latents("examples/female.wav")
 latent_map["AI Beard The Pirate"] = get_latents("examples/pirate_by_coqui.wav")
+latent_map["maid-chan"] = get_latents("examples/remBestPart.mp3")
 
 #### GRADIO INTERFACE ####
 
@@ -746,7 +727,7 @@ EXAMPLES = [
     
 ]
 
-MODELS = ["Zephyr 7B Beta","Mistral 7B Instruct"]
+MODELS = ["Mistral 7B Instruct"]
 
 OTHER_HTML=f"""<div>
 <a style="display:inline-block" href='https://github.com/coqui-ai/TTS'><img src='https://img.shields.io/github/stars/coqui-ai/TTS?style=social' /></a>