LLMs-from-scratch/ch06/01_main-chapter-code/exercise-solutions.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "ba450fb1-8a26-4894-ab7a-5d7bfefe90ce",
   "metadata": {},
   "source": [
    "<table style=\"width:100%\">\n",
    "<tr>\n",
    "<td style=\"vertical-align:middle; text-align:left;\">\n",
    "<font size=\"2\">\n",
    "Supplementary code for the <a href=\"http://mng.bz/orYv\">Build a Large Language Model From Scratch</a> book by <a href=\"https://sebastianraschka.com\">Sebastian Raschka</a><br>\n",
    "<br>Code repository: <a href=\"https://github.com/rasbt/LLMs-from-scratch\">https://github.com/rasbt/LLMs-from-scratch</a>\n",
    "</font>\n",
    "</td>\n",
    "<td style=\"vertical-align:middle; text-align:left;\">\n",
    "<a href=\"http://mng.bz/orYv\"><img src=\"https://sebastianraschka.com/images/LLMs-from-scratch-images/cover-small.webp\" width=\"100px\"></a>\n",
    "</td>\n",
    "</tr>\n",
    "</table>"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "51c9672d-8d0c-470d-ac2d-1271f8ec3f14",
   "metadata": {},
   "source": [
    "# Chapter 6 Exercise solutions"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5fea8be3-30a1-4623-a6d7-b095c6c1092e",
   "metadata": {},
   "source": [
    "## Exercise 6.1: Increasing the context length"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5860ba9f-2db3-4480-b96b-4be1c68981eb",
   "metadata": {},
   "source": [
    "We can pad the inputs to the maximum number of tokens the model supports by setting the max length to 1024:\n",
    "\n",
    "```python\n",
    "max_length = 1024\n",
    "\n",
    "train_dataset = SpamDataset(base_path / \"train.csv\", max_length=max_length, tokenizer=tokenizer)\n",
    "val_dataset = SpamDataset(base_path / \"validation.csv\", max_length=max_length, tokenizer=tokenizer)\n",
    "test_dataset = SpamDataset(base_path / \"test.csv\", max_length=max_length, tokenizer=tokenizer)\n",
    "```\n",
    "\n",
    "or, equivalently, we can define the `max_length` via:\n",
    "\n",
    "```python\n",
    "max_length = model.pos_emb.weight.shape[0]\n",
    "```\n",
    "\n",
    "or\n",
    "\n",
    "```python\n",
    "max_length = BASE_CONFIG[\"context_length\"]\n",
    "```"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2b0f4d5d-17fd-4265-93d8-ea08a22fdaf8",
   "metadata": {},
   "source": [
    "For convenience, you can run this experiment via\n",
    "\n",
    "```bash\n",
    "python additional-experiments.py --context_length \"model_context_length\"\n",
    "```\n",
    "\n",
    "using the code in the [../02_bonus_additional-experiments](../02_bonus_additional-experiments) folder, which results in a substantially worse test accuracy of 78.33% (versus the 95.67% in the main chapter)."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5a780455-f52a-48d1-ab82-6afd40bcad8b",
   "metadata": {},
   "source": [
    "## Exercise 6.2: Finetuning the whole model"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "56aa5208-aa29-4165-a0ec-7480754e2a18",
   "metadata": {},
   "source": [
    "Instead of finetuning just the final transformer block, we can finetune the entire model by removing the following lines from the code:\n",
    "\n",
    "```python\n",
    "for param in model.parameters():\n",
    "    param.requires_grad = False\n",
    "```\n",
    "\n",
    "For convenience, you can run this experiment via\n",
    "\n",
    "```bash\n",
    "python additional-experiments.py --trainable_layers all\n",
    "```\n",
    "\n",
    "using the code in the [../02_bonus_additional-experiments](../02_bonus_additional-experiments) folder, which results in a 1% improved test accuracy of 96.67% (versus the 95.67% in the main chapter)."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2269bce3-f2b5-4a76-a692-5977c75a57b6",
   "metadata": {},
   "source": [
    "## Exercise 6.3: Finetuning the first versus last token "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7418a629-51b6-4aa2-83b7-bc0261bc370f",
   "metadata": {},
   "source": [
    "Rather than finetuning the last output token, we can finetune the first output token by changing \n",
    "\n",
    "```python\n",
    "model(input_batch)[:, -1, :]\n",
    "```\n",
    "\n",
    "to\n",
    "\n",
    "```python\n",
    "model(input_batch)[:, 0, :]\n",
    "```\n",
    "\n",
    "everywhere in the code.\n",
    "\n",
    "For convenience, you can run this experiment via\n",
    "\n",
    "```\n",
    "python additional-experiments.py --trainable_token first\n",
    "```\n",
    "\n",
    "using the code in the [../02_bonus_additional-experiments](../02_bonus_additional-experiments) folder, which results in a substantially worse test accuracy of 75.00% (versus the 95.67% in the main chapter)."
   ]
  }
 ],
 "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": "3.10.11"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}