{
 "cells": [
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    " # Install Dependencies"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "!pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu117 --upgrade"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "!pip install langchain einops accelerate transformers bitsandbytes"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Import Dependencies"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "d:\\YouTube\\6-06-2023 - Falcon\\falcon\\lib\\site-packages\\tqdm\\auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html\n",
      "  from .autonotebook import tqdm as notebook_tqdm\n"
     ]
    }
   ],
   "source": [
    "from langchain import HuggingFacePipeline\n",
    "from langchain import PromptTemplate,  LLMChain\n",
    "from transformers import AutoTokenizer, AutoModelForCausalLM\n",
    "import transformers\n",
    "import os \n",
    "import torch"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "True"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Check if cuda is available \n",
    "torch.cuda.is_available()"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Build the Pipeline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define Model ID\n",
    "model_id = \"tiiuae/falcon-40b-instruct\"\n",
    "# Load Tokenizer\n",
    "tokenizer = AutoTokenizer.from_pretrained(model_id)\n",
    "# Load Model \n",
    "model = AutoModelForCausalLM.from_pretrained(model_id, cache_dir='./workspace/', \n",
    "    torch_dtype=torch.bfloat16, trust_remote_code=True, device_map=\"auto\", offload_folder=\"offload\")\n",
    "# Set PT model to inference mode\n",
    "model.eval()\n",
    "# Build HF Transformers pipeline \n",
    "pipeline = transformers.pipeline(\n",
    "    \"text-generation\", \n",
    "    model=model,\n",
    "    tokenizer=tokenizer,\n",
    "    device_map=\"auto\",\n",
    "    max_length=400,\n",
    "    do_sample=True,\n",
    "    top_k=10,\n",
    "    num_return_sequences=1,\n",
    "    eos_token_id=tokenizer.eos_token_id\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Test out the pipeline\n",
    "pipeline('who is kim kardashian?')"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Pass it to Langchain"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Setup prompt template\n",
    "template = PromptTemplate(input_variables=['input'], template='{input}') \n",
    "# Pass hugging face pipeline to langchain class\n",
    "llm = HuggingFacePipeline(pipeline=pipeline) \n",
    "# Build stacked LLM chain i.e. prompt-formatting + LLM\n",
    "chain = LLMChain(llm=llm, prompt=template)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Test LLMChain \n",
    "response = chain.run('who is kim kardashian?')"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Build Gradio App"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Install Gradio for the UI component\n",
    "!pip install gradio"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Import gradio for UI\n",
    "import gradio as gr"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Create generate function - this will be called when a user runs the gradio app \n",
    "def generate(prompt): \n",
    "    # The prompt will get passed to the LLM Chain!\n",
    "    return chain.run(prompt)\n",
    "    # And will return responses "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define a string variable to hold the title of the app\n",
    "title = '🦜🔗 Falcon-40b-Instruct'\n",
    "# Define another string variable to hold the description of the app\n",
    "description = 'This application demonstrates the use of the open-source `Falcon-40b-Instruct` LLM.'\n",
    "# pls subscribe 🙏"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Build gradio interface, define inputs and outputs...just text in this\n",
    "gr.Interface(fn=generate, inputs=[\"text\"], outputs=[\"text\"], \n",
    "             # Pass through title and description\n",
    "             title=title, description=description, \n",
    "             # Set theme and launch parameters\n",
    "             theme='finlaymacklon/boxy_violet').launch(server_port=8080, share=True)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "falcon",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.12"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}