Upload 6 files

LLM.py

@@ -0,0 +1,101 @@
+from langchain_core.prompts import ChatPromptTemplate, PromptTemplate
+from langchain_groq import ChatGroq
+from langchain_huggingface import ChatHuggingFace
+from langchain_huggingface import HuggingFaceEndpoint
+from dotenv import load_dotenv
+from langchain.schema.output_parser import StrOutputParser
+from langchain_huggingface import ChatHuggingFace
+from langchain_google_genai import ChatGoogleGenerativeAI
+import os
+from huggingface_hub import login
+
+load_dotenv()
+
+login(token=os.environ["HUGGING_FACE_API_KEY"])
+os.environ['CURL_CA_BUNDLE'] = ''
+
+load_dotenv()
+
+class Bot():
+    def __init__(self):
+        self.groq_models = ['gemma-7b-it', 'llama3-70b-8192',\
+                            'llama3-8b-8192', 'mixtral-8x7b-32768']
+        self.hf_models = ["01-ai/Yi-1.5-34B-Chat", "google/gemma-1.1-2b-it",\
+                          "google/gemma-1.1-7b-it"]
+        self.google_models = ["gemini-1.0-pro", "gemini-1.5-flash",\
+                              "gemini-1.5-pro"]
+        self.models = ["gemini-1.0-pro", "gemini-1.5-flash", "gemini-1.5-pro", "01-ai/Yi-1.5-34B-Chat", "google/gemma-1.1-2b-it",\
+                          "google/gemma-1.1-7b-it", 'gemma-7b-it', 'llama3-70b-8192', 'llama3-8b-8192', 'mixtral-8x7b-32768']
+
+    def call_groq(self, model, temp = 0.7, given_prompt = "Hi"):
+        try:
+            llm = ChatGroq(
+                temperature=temp,
+                model= model
+            )
+
+            system = "You are a helpful assistant."
+            human = "{text}"
+            prompt = ChatPromptTemplate.from_messages([("system", system), ("human", human)])
+
+            chain = prompt | llm | StrOutputParser()
+            return chain.invoke({"text": given_prompt})
+
+        except Exception as e:
+            return f"Error: {str(e)}"
+
+    def call_hf(self,model, temp = 0.7, given_prompt = "Hi"):
+        try:
+            llm = HuggingFaceEndpoint(
+                repo_id=model,
+                temperature=temp
+                )
+
+            chat = ChatHuggingFace(llm=llm, verbose=True)
+
+            template = """
+                You are a helpful assistant
+
+                User: {query}
+
+                Answer:
+            """
+
+            prompt = PromptTemplate(
+                template=template,
+                input_variables=["query"]
+            )
+
+            chain =prompt | chat | StrOutputParser()
+
+            return chain.invoke({"query": given_prompt})
+
+        except Exception as e:
+            return f"Error: {str(e)}"
+
+    def call_google(self,model, temp=0.7, given_prompt = "Hi"):
+        try:
+            model = ChatGoogleGenerativeAI(model = model, temprature = temp)
+            system = "You are a helpful assistant."
+            human = "{text}"
+            prompt = ChatPromptTemplate.from_messages([("human", human)])
+            chain = prompt | model | StrOutputParser()
+            return chain.invoke({"text": given_prompt})
+        except Exception as e:
+            return f"Error: {str(e)}"
+
+    def response(self, model, prompt="Hi", temprature = 0.7):
+        if model in self.groq_models:
+            res_show = self.call_groq(temp = temprature, given_prompt = prompt, model= model)
+        elif model in self.hf_models:
+            res_show = self.call_hf(given_prompt = prompt, temp = temprature, model = model)
+        elif model in self.google_models:
+            res_show = self.call_google(given_prompt = prompt, temp = temprature, model = model)
+        else:
+            return "Sorry! App not working properly"
+        return res_show
+
+
+
+
+

app.py

@@ -0,0 +1,111 @@
+import gradio as gr
+from prompt_parser import Parse_Prompt
+from scoreboard import Score
+import warnings
+
+warnings.filterwarnings("ignore")
+
+arena = Parse_Prompt()
+score = Score()
+
+with gr.Blocks(fill_height = True) as app:
+    with gr.Tab("🪖 Battle Field"):
+        gr.Markdown('''## ⚔️ LLM: Large Language Mayhem
+                    - Voting should be fair and based on the performance of the models.
+                    - No cheating or manipulating the outcomes.
+                    - Press 🎲 Random to change the models.
+                    - Everything else except the Random button will only clear the screen, model being the same.
+                    - Have fun and enjoy the language mayhem!
+                    - Warrior names will be visible after your first query (after random also you will have to give a query to see changes)
+                    - Don't See Warrior names before voting
+                    ''')
+        with gr.Row():
+            with gr.Accordion("🥷 Warriors", open = False):
+                gr.Dataframe([[model] for model in arena.models], col_count = 1, headers = ["🥷"])
+        with gr.Group():
+            with gr.Row():
+                with gr.Column():
+                    chatbox1 = gr.Chatbot(label = "Warrior A", show_copy_button = True)
+                with gr.Column():
+                    chatbox2 = gr.Chatbot(label = "Warrior B", show_copy_button = True)
+            textbox = gr.Textbox(show_label = False, placeholder = "👉 Enter your prompt")
+        with gr.Row():
+            with gr.Accordion("🥷 Current Warriors",open = False):
+                with gr.Row():
+                    war1= gr.Textbox(arena.model1, interactive= False, show_label=False)
+                    war2 = gr.Textbox(arena.model2, interactive= False, show_label= False)
+        with gr.Row():
+            with gr.Accordion("👆 Vote", open = False):
+                with gr.Row():
+                    vote_a = gr.ClearButton([textbox, chatbox1, chatbox2], value = "👈 Warrior A Wins")
+                    vote_b = gr.ClearButton([textbox, chatbox1, chatbox2], value = "👉 Warrior B Wins")
+                    vote_tie = gr.ClearButton([textbox, chatbox1, chatbox2], value = "🤝 Both Won")
+
+        submit_button = gr.Button("Submit")
+        with gr.Row():
+            new_round = gr.ClearButton( [textbox, chatbox1, chatbox2], value = "🎲New Round🎲")
+            clear = gr.ClearButton([textbox, chatbox1, chatbox2], value = "🧹 Clear")
+        with gr.Row():
+            with gr.Accordion("🔩 Parameters", open = False):
+                temp_slider = gr.Slider(0,1,value = 0.7, step=0.1, label = "Temprature")
+
+        textbox.submit(
+            fn = arena.gen_output,
+            inputs = [temp_slider, textbox],
+            outputs = [chatbox1, chatbox2]
+        )
+        textbox.submit(
+            fn = arena.current_model2,
+            outputs = war2
+        )
+        textbox.submit(
+            fn = arena.current_model1,
+            outputs = war1
+        )
+        submit_button.click(
+            fn = arena.gen_output,
+            inputs = [temp_slider, textbox],
+            outputs = [chatbox1, chatbox2]
+        )
+        submit_button.click(
+            fn = arena.current_model1,
+            outputs = war1
+        )
+        submit_button.click(
+            fn = arena.current_model2,
+            outputs = war2
+        )
+        vote_a.click(
+            fn=lambda: score.update(arena.model1, score.df)
+        )
+        vote_b.click(
+            fn = lambda: score.update(arena.model2, score.df)
+        )
+        vote_tie.click(
+            fn = arena.change_models
+        )
+        new_round.click(
+            fn = arena.change_models
+        )
+        clear.click(
+            fn = arena.clear_history
+        )
+
+    with gr.Tab("💯 Score Board") as data_tab:
+        gr.Markdown('''## ⚔️ LLM: Large Language Mayhem
+                    - Voting should be fair and based on the performance of the models.
+                    - No cheating or manipulating the outcomes.
+                    - Click on Generate button to Update the 💯 Scoreboard.
+                    ''')
+        gr.Interface(
+            fn = score.df_show,
+            inputs = None,
+            outputs=gr.Dataframe(type="pandas", label="Scoreboard", headers = ["","",""]),
+            live = True,
+            allow_flagging = "never",
+            clear_btn = None
+
+        )
+        
+
+app.launch()

prompt_parser.py

@@ -0,0 +1,35 @@
+from LLM import Bot
+import random
+
+class Parse_Prompt(Bot):
+    def __init__(self):
+        super().__init__()
+        self.change = True
+        self.model1 = None
+        self.model2 = None
+        self.chat_history_1 = []
+        self.chat_history_2 = []
+
+    def model_init(self):
+        return random.sample(self.models, 2)
+
+    def clear_history(self):
+        self.chat_history_1 = []
+        self.chat_history_2 = []
+
+    def change_models(self):
+        self.clear_history()
+        self.change = True
+
+    def current_model1(self):
+        return self.model1
+    def current_model2(self):
+        return self.model2
+
+    def gen_output(self, temp, prompt):
+        if self.change:
+            [self.model1, self.model2] = self.model_init()
+            self.change = False
+        self.chat_history_1.append([prompt, self.response(self.model1, prompt, temp)])
+        self.chat_history_2.append([prompt, self.response(self.model2, prompt, temp)])
+        return self.chat_history_1, self.chat_history_2

requirements.txt

@@ -0,0 +1,14 @@
+langchain==0.2.6
+langchain-community==0.2.6
+langchain-core==0.2.10
+langchain-google-genai==1.0.7
+python-dotenv
+langchain-groq==0.1.5
+langchain-huggingface==0.0.3
+langchain-text-splitters==0.2.1
+requests==2.27.1
+numpy
+pandas
+gradio
+huggingface_hub
+sentencepiece

scoreboard.csv

@@ -0,0 +1,11 @@
+Models,Fights Won
+gemini-1.0-pro,2
+gemini-1.5-flash,0
+gemini-1.5-pro,1
+01-ai/Yi-1.5-34B-Chat,0
+google/gemma-1.1-2b-it,5
+google/gemma-1.1-7b-it,0
+gemma-7b-it,0
+llama3-70b-8192,0
+llama3-8b-8192,2
+mixtral-8x7b-32768,1

scoreboard.py

@@ -0,0 +1,28 @@
+import numpy as np
+from prompt_parser import Parse_Prompt
+import pandas as pd
+
+class Score(Parse_Prompt):
+    def __init__(self):
+        super().__init__()
+        self.file_path = 'scoreboard.csv'
+        self.init_scores()
+
+    def init_scores(self):
+        try:
+            self.df = pd.read_csv(self.file_path)
+        except FileNotFoundError:
+            data = {
+                'Models': self.models,
+                'Fights Won': np.zeros(10, dtype = int)
+            }
+            self.df = pd.DataFrame(data)
+            self.df.to_csv(self.file_path, index=False)
+
+    def update(self, model, df):
+        df.loc[self.df["Models"] == model, 'Fights Won'] += 1
+        df.to_csv(self.file_path, index=False)
+        self.clear_history()
+
+    def df_show(self):
+        return pd.read_csv(self.file_path)