Jupyter Notebook Experiment: Trying out orca_mini_v2_13b LLM

orca_experiments.ipynb

{
 "cells": [
  {
   "attachments": {},
   "cell_type": "markdown",
   "id": "36653f03",
   "metadata": {},
   "source": [
    "Trying out [orca_mini_v2_13b/GGML](https://huggingface.co/TheBloke/orca_mini_v2_13b-GGML). Downloaded as follows:\n",
    "```sh\n",
    "python download-model.py --output . --select=\"orca_mini_v2_13b.ggmlv3.q4_K_M.bin|orca_mini_v2_13b.ggmlv3.q5_K_M.bin\" TheBloke/orca_mini_v2_13b-GGML\n",
    "```\n",
    "\n",
    "Model card suggests the following prompt template:\n",
    "\n",
    "```\n",
    "### System:\n",
    "You are an AI assistant that follows instruction extremely well. Help as much as you can.\n",
    "\n",
    "### User:\n",
    "prompt\n",
    "\n",
    "### Input:\n",
    "input, if required\n",
    "\n",
    "### Response:\n",
    "```\n",
    "\n",
    "Which is I think is its own style(?). Of course there are plenty of examples in the [training data](https://huggingface.co/datasets/psmathur/orca_minis_uncensored_dataset)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "6c8fda5c",
   "metadata": {},
   "outputs": [],
   "source": [
    "from functools import partial\n",
    "from ogbujipt.config import openai_emulation\n",
    "from ogbujipt.prompting.basic import context_build\n",
    "from ogbujipt.prompting.model_style import ORCA_DELIMITERS\n",
    "\n",
    "HOST, PORT = 'http://localhost', '8000'\n",
    "\n",
    "openai_api = openai_emulation(host=HOST, port=PORT)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "1f9abcfd",
   "metadata": {},
   "outputs": [],
   "source": [
    "request = partial(openai_api.Completion.create, model='', temperature=0.1, max_tokens=100)\n",
    "\n",
    "BAD_XML_CODE = '''\\\n",
    "<earth>\n",
    "<country><b>Russia</country></b>\n",
    "<capital>Moscow</capital>\n",
    "</Earth>'''\n",
    "\n",
    "prompt = context_build(\n",
    "    'Correct the given XML to make it well-formed',\n",
    "    contexts=BAD_XML_CODE,\n",
    "    delimiters=ORCA_DELIMITERS)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "a79b97d0",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n",
      "<earth>\n",
      "  <country>\n",
      "    <b>Russia</b>\n",
      "  </country>\n",
      "  <capital>Moscow</capital>\n",
      "</earth>\n"
     ]
    }
   ],
   "source": [
    "response = request(prompt=prompt)\n",
    "print(response['choices'][0]['text'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "8a068365",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.11.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}