Update app.py

app.py

@@ -1,130 +1,77 @@
-import optimum
+import os
+import math
 import transformers
-from transformers import AutoConfig, AutoTokenizer, AutoModel, AutoModelForCausalLM
-from optimum.bettertransformer import BetterTransformer
+from transformers import AutoModelForCausalLM, AutoTokenizer
 import torch
 import gradio as gr
-import json
-import os
-import shutil
-import requests
-
-# Define the device
-device = "cuda" if torch.cuda.is_available() else "cpu"
-#Define variables 
-temperature=0.4
-max_new_tokens=240
-top_p=0.92
-repetition_penalty=1.7
-
-model_name = "OpenLLM-France/Claire-7B-0.1"
-
-tokenizer = transformers.AutoTokenizer.from_pretrained(model_name)
-model = transformers.AutoModelForCausalLM.from_pretrained(model_name,
-    device_map="auto",
-    torch_dtype=torch.bfloat16,
-    load_in_4bit=True                          # For efficient inference, if supported by the GPU card
-)
-model = BetterTransformer.transform(model)
-
-
-# Class to encapsulate the Falcon chatbot
-class FalconChatBot:
-    def __init__(self, system_prompt="Le dialogue suivant est une conversation"):
-        self.system_prompt = system_prompt
-
-    def predict(self, user_message, assistant_message, temperature=0.4, max_new_tokens=700, top_p=0.99, repetition_penalty=1.9):
-        # Combine the user and assistant messages into a conversation
-        conversation = f"{self.system_prompt} {assistant_message if assistant_message else ''} {user_message} "
-        # Encode the conversation using the tokenizer
-        input_ids = tokenizer.encode(conversation, return_tensors="pt", add_special_tokens=False)
-        input_ids = input_ids.to(device)
-        # Generate a response using the Falcon model
-        response = model.generate(
-            input_ids=input_ids, 
-            use_cache=False, 
-            early_stopping=False, 
-            bos_token_id=model.config.bos_token_id, 
-            eos_token_id=model.config.eos_token_id, 
-            pad_token_id=model.config.eos_token_id, 
-            temperature=temperature, 
-            do_sample=True, 
-            max_new_tokens=max_new_tokens, 
-            top_p=top_p, 
-            repetition_penalty=repetition_penalty
-        )
-        # Decode the generated response to text
-        response_text = tokenizer.decode(response[0], skip_special_tokens=True)
-        return response_text
 
-# Create the Falcon chatbot instance
-falcon_bot = FalconChatBot()
 
 # Define the Gradio interface
-title = "👋🏻Bienvenue à Tonic's 🌜🌚Claire Chat !"
-description = "Vous pouvez utiliser [🌜🌚ClaireGPT](https://huggingface.co/OpenLLM-France/Claire-7B-0.1) Ou dupliquer pour l'uiliser localement ou sur huggingface! [Join me on Discord to build together](https://discord.gg/VqTxc76K3u)."
-examples = [
-    [
-        "Le dialogue suivant est une conversation entre Emmanuel Macron et Elon Musk:",  # user_message
-        "[Emmanuel Macron]: Bonjour Monsieur Musk. Je vous remercie de me recevoir aujourd'hui.",  # assistant_message
-        0.9,  # temperature
-        150,  # max_new_tokens
-        0.90,  # top_p
-        1.9,  # repetition_penalty
-    ]
-]
-
-additional_inputs=[
-    gr.Textbox("", label="Introduisez Un Personnage Ici ou Mettez En Scene"),
-    gr.Slider(
-        label="Max new tokens",
-        value=100,  # Default value
-        minimum=25,
-        maximum=256,
-        step=1,
-        interactive=True,
-        info="The maximum numbers of new tokens",
-    ),
-    gr.Slider(
-        label="Temperature",
-        value=0.7,  # Default value
-        minimum=0.05,
-        maximum=1.0,
-        step=0.05,
-        interactive=True,
-        info="Higher values produce more diverse outputs",
-    ),
-    gr.Slider(
-        label="Top-p (nucleus sampling)",
-        value=0.90,
-        minimum=0.01,
-        maximum=0.99,
-        step=0.05,
-        interactive=True,
-        info="Higher values sample more low-probability tokens",
-    ),
-    gr.Slider(
-        label="Repetition penalty",
-        value=1.9,
-        minimum=1.0,
-        maximum=2.0,
-        step=0.05,
-        interactive=True,
-        info="Penalize repeated tokens",
-    )
-]
+title = "Welcome to Tonic's 🐋🐳Orca-2-13B!"
+description = "You can use [🐋🐳microsoft/Orca-2-13b](https://huggingface.co/microsoft/Orca-2-13b) Or clone this space to use it locally or on huggingface! [Join me on Discord to build together](https://discord.gg/VqTxc76K3u)."
+
+
+# Load the model and tokenizer
+model_name = "microsoft/Orca-2-13b"
+model = AutoModelForCausalLM.from_pretrained(model_name, device_map='auto')
+tokenizer = AutoTokenizer.from_pretrained(model_name, use_fast=False,)
+
+
+class OrcaChatBot:
+    def __init__(self, model, tokenizer, system_message="You are Orca, an AI language model created by Microsoft. You are a cautious assistant. You carefully follow instructions. You are helpful and harmless and you follow ethical guidelines and promote positive behavior."):
+        self.model = model
+        self.tokenizer = tokenizer
+        self.system_message = system_message
+        self.conversation_history = None
+
+    def predict(self, user_message, temperature=0.4, max_new_tokens=70, top_p=0.99, repetition_penalty=1.9):
+        # Prepare the prompt
+        prompt = f"<|im_start|>system\n{self.system_message}<|im_end|>\n<|im_start|>user\n{user_message}<|im_end|>\n<|im_start|>assistant" if self.conversation_history is None else self.conversation_history + f"<|im_end|>\n<|im_start|>user\n{user_message}<|im_end|>\n<|im_start|>assistant"
+
+        # Encode the prompt
+        inputs = self.tokenizer(prompt, return_tensors='pt', add_special_tokens=False)
+        input_ids = inputs["input_ids"].to(self.model.device)
+
+        # Generate a response
+        output_ids = self.model.generate(
+            input_ids,
+            max_length=input_ids.shape[1] + max_new_tokens,
+            temperature=temperature,
+            top_p=top_p,
+            repetition_penalty=repetition_penalty,
+            pad_token_id=self.tokenizer.eos_token_id
+        )
+
+        # Decode the generated response
+        response = self.tokenizer.decode(output_ids[0], skip_special_tokens=True)
+
+        # Update conversation history
+        self.conversation_history = self.tokenizer.decode(output_ids[0], skip_special_tokens=False)
+
+        return response
+
+Orca_bot = OrcaChatBot(model, tokenizer)
+
+def gradio_predict(user_message, character_intro, max_new_tokens, temperature, top_p, repetition_penalty):
+    # Prepend the character introduction to the user message if provided
+    full_message = f"{system_message}\n{user_message}" if system_message else user_message
+    return Orca_bot.predict(full_message, temperature, max_new_tokens, top_p, repetition_penalty)
 
 iface = gr.Interface(
-    fn=falcon_bot.predict,
+    fn=gradio_predict,
     title=title,
     description=description,
-    examples=examples,
     inputs=[
-        gr.Textbox(label="Utilisez se format pour initier une conversation [Personage:]", type="text", lines=5),
-    ] + additional_inputs,
+        gr.Textbox(label="Your Message", type="text", lines=3),
+        gr.Textbox(label="Introduce a Character Here or Set a Scene (system prompt)", type="text", lines=2),
+        gr.Slider(label="Max new tokens", value=1200, minimum=25, maximum=4096, step=1),
+        gr.Slider(label="Temperature", value=0.7, minimum=0.05, maximum=1.0, step=0.05),
+        gr.Slider(label="Top-p (nucleus sampling)", value=0.90, minimum=0.01, maximum=0.99, step=0.05),
+        gr.Slider(label="Repetition penalty", value=1.9, minimum=1.0, maximum=2.0, step=0.05)
+    ],
     outputs="text",
     theme="ParityError/Anime"
 )
 
-# Launch the Gradio interface for the Falcon model
+# Launch the Gradio interface
 iface.launch()