Lesson_2-Batching.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "239be6af",
   "metadata": {},
   "source": [
    "# Lesson 2 - Batching\n",
    "\n",
    "In this lesson, we'll discuss the concept of \"batching\" in LLM inference.\n",
    "\n",
    "- What is batching?\n",
    "- Throughput vs latency"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0dd87319-4656-45f1-a6be-0a0b970e01f9",
   "metadata": {},
   "source": [
    "### Import required packages and load the LLM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "3ea434a9-22e6-4bee-8ef3-65e8992ce089",
   "metadata": {
    "height": 114
   },
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import time\n",
    "import torch\n",
    "from transformers import AutoModelForCausalLM, AutoTokenizer\n",
    "from tqdm import tqdm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "7066fad6-94df-4153-a6de-a33d7dace6b4",
   "metadata": {
    "height": 63
   },
   "outputs": [],
   "source": [
    "model_name = \"./models/gpt2\"\n",
    "tokenizer = AutoTokenizer.from_pretrained(model_name)\n",
    "model = AutoModelForCausalLM.from_pretrained(model_name)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f6ca61e7-ad31-49ae-9338-6e38b0881dde",
   "metadata": {},
   "source": [
    "### Reuse KV-cache text generation function from Lesson 1\n",
    "- Use the same prompt as the previous lesson to verify everything is working as expected"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "a1b79f19-38f6-4f34-804d-921ceb40931b",
   "metadata": {
    "height": 636
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " fence and ran to the other side of the fence\n"
     ]
    }
   ],
   "source": [
    "prompt = \"The quick brown fox jumped over the\"\n",
    "inputs = tokenizer(prompt, return_tensors=\"pt\")\n",
    "\n",
    "\n",
    "def generate_token_with_past(inputs):\n",
    "    with torch.no_grad():\n",
    "        outputs = model(**inputs)\n",
    "\n",
    "    logits = outputs.logits\n",
    "    last_logits = logits[0, -1, :]\n",
    "    next_token_id = last_logits.argmax()\n",
    "    return next_token_id, outputs.past_key_values\n",
    "\n",
    "\n",
    "def generate(inputs, max_tokens):\n",
    "    generated_tokens = []\n",
    "    next_inputs = inputs\n",
    "    for _ in range(max_tokens):\n",
    "        next_token_id, past_key_values = \\\n",
    "        generate_token_with_past(next_inputs)\n",
    "        next_inputs = {\n",
    "            \"input_ids\": next_token_id.reshape((1, 1)),\n",
    "            \"attention_mask\": torch.cat(\n",
    "                [next_inputs[\"attention_mask\"], torch.tensor([[1]])],\n",
    "                dim=1\n",
    "            ),\n",
    "            \"past_key_values\": past_key_values,\n",
    "        }\n",
    "\n",
    "        next_token = tokenizer.decode(next_token_id)\n",
    "        generated_tokens.append(next_token)\n",
    "    return \"\".join(generated_tokens)\n",
    "\n",
    "\n",
    "tokens = generate(inputs, max_tokens=10)\n",
    "print(tokens)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "802ad913-0c0e-450b-afa0-ff289b85b5f1",
   "metadata": {},
   "source": [
    "### Add padding tokens to the model to prepare batches of prompts"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "d895c2e9-fd7f-41d9-af00-a44a19d512b2",
   "metadata": {
    "height": 63
   },
   "outputs": [],
   "source": [
    "# Define PAD Token = EOS Token = 50256\n",
    "tokenizer.pad_token = tokenizer.eos_token\n",
    "model.config.pad_token_id = model.config.eos_token_id"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "e09245a5-d032-4a45-a2e8-dad82305b449",
   "metadata": {
    "height": 63
   },
   "outputs": [],
   "source": [
    "# pad on the left so we can append new tokens on the right\n",
    "tokenizer.padding_side = \"left\"\n",
    "tokenizer.truncation_side = \"left\""
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0ea8d58c-3732-4b00-9819-9d3c45e4fd04",
   "metadata": {},
   "source": [
    "- Tokenize list of prompts\n",
    "- Add padding so that all prompts have the same number of tokens as the longest prompt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "17dc969e-eaf1-4594-b328-2fbb8dbd470a",
   "metadata": {
    "height": 199
   },
   "outputs": [],
   "source": [
    "# multiple prompts of varying lengths to send\n",
    "# to the model at once\n",
    "prompts = [\n",
    "    \"The quick brown fox jumped over the\",\n",
    "    \"The rain in Spain falls\",\n",
    "    \"What comes up must\",\n",
    "]\n",
    "\n",
    "# note: padding=True ensures the padding token\n",
    "# will be inserted into the tokenized tensors\n",
    "inputs = tokenizer(prompts, padding=True, return_tensors=\"pt\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "54650f21-92cd-4d78-b2e3-ddd3dd811687",
   "metadata": {
    "height": 46
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "input_ids: tensor([[  464,  2068,  7586, 21831, 11687,   625,   262],\n",
      "        [50256, 50256,   464,  6290,   287,  8602,  8953],\n",
      "        [50256, 50256, 50256,  2061,  2058,   510,  1276]])\n",
      "shape: torch.Size([3, 7])\n"
     ]
    }
   ],
   "source": [
    "print(\"input_ids:\", inputs[\"input_ids\"])\n",
    "print(\"shape:\", inputs[\"input_ids\"].shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "f80e4eaa-819e-4209-9154-58b77464043d",
   "metadata": {
    "height": 46
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "attention_mask: tensor([[1, 1, 1, 1, 1, 1, 1],\n",
      "        [0, 0, 1, 1, 1, 1, 1],\n",
      "        [0, 0, 0, 1, 1, 1, 1]])\n",
      "shape: torch.Size([3, 7])\n"
     ]
    }
   ],
   "source": [
    "print(\"attention_mask:\", inputs[\"attention_mask\"])\n",
    "print(\"shape:\", inputs[\"attention_mask\"].shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5264a70b-d66c-439a-873b-732579eb71c3",
   "metadata": {},
   "source": [
    "- Add position ids to track original order of tokens in each prompt\n",
    "- Padding tokens are set to `1` and then first real token starts with position `0`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "5a9fd5ca-7bc0-48ad-a6f4-88c29598570c",
   "metadata": {
    "height": 148
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[0, 1, 2, 3, 4, 5, 6],\n",
       "        [1, 1, 0, 1, 2, 3, 4],\n",
       "        [1, 1, 1, 0, 1, 2, 3]])"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# position_ids tell the transformer the ordinal position\n",
    "# of each token in the input sequence\n",
    "# for single input inference, this is just [0 .. n]\n",
    "# for n tokens, but for batch inference,\n",
    "# we need to 0 out the padding tokens at the start of the sequence\n",
    "attention_mask = inputs[\"attention_mask\"]\n",
    "position_ids = attention_mask.long().cumsum(-1) - 1\n",
    "position_ids.masked_fill_(attention_mask == 0, 1)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3d68e22e-73a2-4527-9529-11cf98ea139e",
   "metadata": {},
   "source": [
    "- Pass tokens to model to calculate logits"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "117ed42c-cc0b-4ca8-b656-dd1484fc5574",
   "metadata": {
    "height": 80
   },
   "outputs": [],
   "source": [
    "# same as before, but include the position_ids\n",
    "with torch.no_grad():\n",
    "    outputs = model(position_ids=position_ids, **inputs)\n",
    "logits = outputs.logits"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2cb19b85-673a-4f0c-ac11-3e82b0f9ddc9",
   "metadata": {},
   "source": [
    "- Retrieve most likely token for each prompt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "a70df683-bc0a-403b-9cbf-014cb2362f01",
   "metadata": {
    "height": 46
   },
   "outputs": [],
   "source": [
    "last_logits = logits[:, -1, :] \n",
    "next_token_ids = last_logits.argmax(dim=1) "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ede5ecf0",
   "metadata": {},
   "source": [
    "- Print the next token ids"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "74f61f3a-5f19-47e4-92cc-8d30ef574944",
   "metadata": {
    "height": 29
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor([13990,   319,   307])\n"
     ]
    }
   ],
   "source": [
    "print(next_token_ids)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7c645c96",
   "metadata": {},
   "source": [
    "- Convert the token ids into strings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "78a8061b-643f-4125-a5a7-fb6461523dfe",
   "metadata": {
    "height": 46
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[' fence', ' on', ' be']"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "next_tokens = tokenizer.batch_decode(next_token_ids)\n",
    "next_tokens"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ad750c23",
   "metadata": {},
   "source": [
    "### Let's put it all together!\n",
    " - Generate n tokens with past"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "d3543bb0-2a33-49dc-9d61-5af911a03d24",
   "metadata": {
    "height": 148
   },
   "outputs": [],
   "source": [
    "def generate_batch_tokens_with_past(inputs):\n",
    "    with torch.no_grad():\n",
    "        outputs = model(**inputs)\n",
    "\n",
    "    logits = outputs.logits\n",
    "    last_logits = logits[:, -1, :]\n",
    "    next_token_ids = last_logits.argmax(dim=1)\n",
    "    return next_token_ids, outputs.past_key_values"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ea81cfb2",
   "metadata": {},
   "source": [
    "- Generate all tokens for some max tokens"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "bb50b1e6-415f-4315-8816-8818df516a07",
   "metadata": {
    "height": 585
   },
   "outputs": [],
   "source": [
    "def generate_batch(inputs, max_tokens):\n",
    "    # create a list of tokens for every input in the batch\n",
    "    generated_tokens = [\n",
    "        [] for _ in range(inputs[\"input_ids\"].shape[0])\n",
    "    ]\n",
    "\n",
    "    attention_mask = inputs[\"attention_mask\"]\n",
    "    position_ids = attention_mask.long().cumsum(-1) - 1\n",
    "    position_ids.masked_fill_(attention_mask == 0, 1)\n",
    "\n",
    "    next_inputs = {\n",
    "        \"position_ids\": position_ids,\n",
    "        **inputs\n",
    "    }\n",
    "\n",
    "    for _ in range(max_tokens):\n",
    "        next_token_ids, past_key_values = \\\n",
    "            generate_batch_tokens_with_past(next_inputs)\n",
    "\n",
    "        next_inputs = {\n",
    "            \"input_ids\": next_token_ids.reshape((-1, 1)),\n",
    "            \"position_ids\": next_inputs[\"position_ids\"][:, -1].unsqueeze(-1) + 1,\n",
    "            \"attention_mask\": torch.cat([\n",
    "                next_inputs[\"attention_mask\"],\n",
    "                torch.ones((next_token_ids.shape[0], 1)),  \n",
    "            ], dim=1),\n",
    "            \"past_key_values\": past_key_values,\n",
    "        }\n",
    "\n",
    "        next_tokens = tokenizer.batch_decode(next_token_ids)\n",
    "        for i, token in enumerate(next_tokens):\n",
    "            generated_tokens[i].append(token)\n",
    "    return [\"\".join(tokens) for tokens in generated_tokens]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6c8845f1",
   "metadata": {},
   "source": [
    "- Call the generate_batch function and print out the generated tokens"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "4bd8d5cc-c06c-4efc-97da-ebcd3b2e15b6",
   "metadata": {
    "height": 29
   },
   "outputs": [],
   "source": [
    "generated_tokens = generate_batch(inputs, max_tokens=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "aa48f7c0-6633-400e-aa11-e74000497b65",
   "metadata": {
    "height": 46
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The quick brown fox jumped over the \u001b[31m fence and ran to the other side of the fence\u001b[0m\n",
      "\n",
      "The rain in Spain falls \u001b[31m on the first day of the month, and the\u001b[0m\n",
      "\n",
      "What comes up must \u001b[31m be a good idea.\n",
      "\n",
      "\"I think\u001b[0m\n",
      "\n"
     ]
    }
   ],
   "source": [
    "for prompt, generated in zip(prompts, generated_tokens):\n",
    "    print(prompt, f\"\\x1b[31m{generated}\\x1b[0m\\n\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e077e2c0",
   "metadata": {},
   "source": [
    "### Throughput vs Latency\n",
    "\n",
    "- Explore the effect of batching on latency (how long it takes to generate each token). \n",
    "- Observe the fundamental tradeoff that exists between throughput and latency.\n",
    "\n",
    "**Note:** Your results might differ somewhat from those shown in the video, but they will still follow the same pattern as explained by the instructor."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "35e3a944-e2f3-4cda-a5e3-d753b223e73a",
   "metadata": {
    "height": 602
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "bs= 1\n",
      "duration 1.737901210784912\n",
      "throughput 5.754066996410897\n",
      "avg latency 0.1737901210784912\n",
      "\n",
      "bs= 2\n",
      "duration 2.2069575786590576\n",
      "throughput 9.062249403158875\n",
      "avg latency 0.22069575786590576\n",
      "\n",
      "bs= 4\n",
      "duration 2.589951992034912\n",
      "throughput 15.444301717952774\n",
      "avg latency 0.25899519920349123\n",
      "\n",
      "bs= 8\n",
      "duration 2.6060476303100586\n",
      "throughput 30.697827265145524\n",
      "avg latency 0.2606047630310059\n",
      "\n",
      "bs= 16\n",
      "duration 3.497083902359009\n",
      "throughput 45.75240528031646\n",
      "avg latency 0.3497083902359009\n",
      "\n",
      "bs= 32\n",
      "duration 3.502141237258911\n",
      "throughput 91.37267126623956\n",
      "avg latency 0.3502141237258911\n",
      "\n",
      "bs= 64\n",
      "duration 4.898610591888428\n",
      "throughput 130.64929085397625\n",
      "avg latency 0.48986105918884276\n",
      "\n",
      "bs= 128\n",
      "duration 8.49881362915039\n",
      "throughput 150.6092562860399\n",
      "avg latency 0.849881362915039\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# constants\n",
    "max_tokens = 10\n",
    "\n",
    "# observations\n",
    "durations = []\n",
    "throughputs = []\n",
    "latencies = []\n",
    "\n",
    "batch_sizes = [2**p for p in range(8)]\n",
    "for batch_size in batch_sizes:\n",
    "    print(f\"bs= {batch_size}\")\n",
    "\n",
    "    # generate tokens for batch and record duration\n",
    "    t0 = time.time()\n",
    "    batch_prompts = [\n",
    "        prompts[i % len(prompts)] for i in range(batch_size)\n",
    "    ]\n",
    "    inputs = tokenizer(\n",
    "        batch_prompts, padding=True, return_tensors=\"pt\"\n",
    "    )\n",
    "    generated_tokens = generate_batch(inputs, max_tokens=max_tokens)\n",
    "    duration_s = time.time() - t0\n",
    "\n",
    "    ntokens = batch_size * max_tokens\n",
    "    throughput = ntokens / duration_s\n",
    "    avg_latency = duration_s / max_tokens\n",
    "    print(\"duration\", duration_s)\n",
    "    print(\"throughput\", throughput)\n",
    "    print(\"avg latency\", avg_latency)    \n",
    "    print()\n",
    "\n",
    "    durations.append(duration_s)\n",
    "    throughputs.append(throughput)\n",
    "    latencies.append(avg_latency)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5ff34a13",
   "metadata": {},
   "source": [
    "### Let's plot the throughput and latency observations against the batch size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "898c44fc-e5ba-48dd-8d80-aae9eddec98b",
   "metadata": {
    "height": 398
   },
   "outputs": [],
   "source": [
    "def render_plot(x, y1, y2, x_label, y1_label, y2_label):\n",
    "    # Create a figure and a set of subplots\n",
    "    fig, ax1 = plt.subplots()\n",
    "\n",
    "    # Plot the first line (throughput)\n",
    "    color = 'tab:red'\n",
    "    ax1.set_xlabel(x_label)\n",
    "    ax1.set_ylabel(y1_label, color=color)\n",
    "    ax1.plot(x, y1, color=color)\n",
    "    ax1.tick_params(axis='y', labelcolor=color)\n",
    "\n",
    "    # Set the x-axis to be log-scaled\n",
    "    ax1.set_xscale('log', base=2)\n",
    "\n",
    "    # Instantiate a second axes that shares the same x-axis\n",
    "    ax2 = ax1.twinx()  \n",
    "    color = 'tab:blue'\n",
    "    ax2.set_ylabel(y2_label, color=color)  # we already handled the x-label with ax1\n",
    "    ax2.plot(x, y2, color=color)\n",
    "    ax2.tick_params(axis='y', labelcolor=color)\n",
    "\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c6c99edb",
   "metadata": {},
   "source": [
    "**Note**: Your plot may vary slightly from the one shown in the video, yet it will exhibit a similar pattern."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "01e2f578-bb58-404a-9611-79b7cf79ed09",
   "metadata": {
    "height": 148
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "render_plot(\n",
    "    batch_sizes,\n",
    "    throughputs,\n",
    "    latencies,\n",
    "    \"Batch Size\",\n",
    "    \"Throughput\",\n",
    "    \"Latency\"\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "3d650c2f",
   "metadata": {
    "height": 29
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "8537b718",
   "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": "3.10.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}