llama-13b-qlora-remote-decorator_selfsupervised.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Fine-tune Llama-13B with QLoRA and SageMaker remote decorator\n",
    "\n",
    "## Self-supervised fine-tuning"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "\n",
    "In this demo notebook, we demonstrate how to fine-tune the Llama-13B model using QLoRA, Hugging Face PEFT, and bitsandbytes.\n",
    "\n",
    "We are using SageMaker remote decorator for runinng the fine-tuning job on Amazon SageMaker Training job\n",
    "---\n",
    "\n",
    "JupyterLab Instance Type: ml.t3.medium\n",
    "\n",
    "SageMaker Distribution image: SageMaker Distribution 2.0\n",
    "\n",
    "PyTorch version: 3.11\n",
    "\n",
    "Fine-Tuning:\n",
    "* Instance Type: ml.g5.12xlarge"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Install the required libriaries, including the Hugging Face libraries, and restart the kernel."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-07-20T12:38:06.851473Z",
     "start_time": "2023-07-20T12:38:04.440644Z"
    },
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "%pip install -r requirements.txt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-09-07T11:36:57.011485Z",
     "start_time": "2023-09-07T11:36:36.228753Z"
    },
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "%pip install -q -U langchain==0.1.5"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\n",
    "## Setup Configuration file path\n",
    "\n",
    "We are setting the directory in which the config.yaml file resides so that remote decorator can make use of the settings through [SageMaker Defaults](https://sagemaker.readthedocs.io/en/stable/overview.html#configuring-and-using-defaults-with-the-sagemaker-python-sdk).\n",
    "\n",
    "This notebook is using the Hugging Face container for the `us-east-1` region. Make sure you are using the right image for your AWS region, otherwise edit [config.yaml](./config.yaml). Container Images are available [here](https://github.com/aws/deep-learning-containers/blob/master/available_images.md)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-11-15T09:25:13.511301Z",
     "start_time": "2023-11-15T09:25:13.502916Z"
    },
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "import os\n",
    "\n",
    "# Set path to config file\n",
    "os.environ[\"SAGEMAKER_USER_CONFIG_OVERRIDE\"] = os.getcwd()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Visualize and upload the dataset\n",
    "\n",
    "Read train dataset in a Pandas dataframe"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-11-15T09:25:48.236440Z",
     "start_time": "2023-11-15T09:25:24.568506Z"
    },
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "from langchain.document_loaders import WebBaseLoader\n",
    "\n",
    "loader = WebBaseLoader([\n",
    "    \"https://aws.amazon.com/bedrock/\",\n",
    "    \"https://docs.aws.amazon.com/bedrock/latest/userguide/what-is-bedrock.html\",\n",
    "    \"https://aws.amazon.com/blogs/aws/preview-enable-foundation-models-to-complete-tasks-with-agents-for-amazon-bedrock/\",\n",
    "    \"https://docs.aws.amazon.com/bedrock/latest/userguide/agents.html\",\n",
    "    \"https://docs.aws.amazon.com/bedrock/latest/userguide/knowledge-base.html\",\n",
    "\n",
    "])\n",
    "\n",
    "data = loader.load()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-11-15T09:26:19.870311Z",
     "start_time": "2023-11-15T09:26:19.017194Z"
    },
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    }
   },
   "outputs": [],
   "source": [
    "from datasets import Dataset\n",
    "\n",
    "def strip_spaces(doc):\n",
    "    return {\"text\": doc.page_content.replace(\"  \", \"\")}\n",
    "\n",
    "stripped_data = list(map(strip_spaces, data))\n",
    "\n",
    "train_dataset = Dataset.from_list(stripped_data)\n",
    "\n",
    "train_dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Use the Hugging Face Trainer class to fine-tune the model. Define the hyperparameters we want to use. We also create a DataCollator that will take care of padding our inputs and labels."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-11-15T09:26:25.674063Z",
     "start_time": "2023-11-15T09:26:25.154338Z"
    },
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "! huggingface-cli login --token <HF_TOKEN>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-11-15T09:26:30.744432Z",
     "start_time": "2023-11-15T09:26:26.861979Z"
    },
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "from transformers import AutoTokenizer\n",
    "\n",
    "model_id = \"meta-llama/Llama-2-13b-chat-hf\"\n",
    "\n",
    "tokenizer = AutoTokenizer.from_pretrained(model_id, use_auth_token=True)\n",
    "\n",
    "tokenizer.pad_token = tokenizer.eos_token"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Creating chunks and tokenizing the inputs for making it usable by the LLM. For additional details, please refer to the blog [Leveraging qLoRA for Fine-Tuning of Task-Fine-Tuned Models Without Catastrophic Forgetting: A Case Study with LLaMA2(-chat)](https://medium.com/towards-data-science/leveraging-qlora-for-fine-tuning-of-task-fine-tuned-models-without-catastrophic-forgetting-d9bcd594cff4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-11-15T09:26:37.996087Z",
     "start_time": "2023-11-15T09:26:37.991057Z"
    },
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    }
   },
   "outputs": [],
   "source": [
    "from itertools import chain\n",
    "from functools import partial\n",
    "\n",
    "remainder = {\"input_ids\": [], \"attention_mask\": [], \"token_type_ids\": []}\n",
    "\n",
    "def chunk(sample, chunk_length=2048):\n",
    "    # define global remainder variable to save remainder from batches to use in next batch\n",
    "    global remainder\n",
    "    # Concatenate all texts and add remainder from previous batch\n",
    "    concatenated_examples = {k: list(chain(*sample[k])) for k in sample.keys()}\n",
    "    concatenated_examples = {k: remainder[k] + concatenated_examples[k] for k in concatenated_examples.keys()}\n",
    "    # get total number of tokens for batch\n",
    "    batch_total_length = len(concatenated_examples[list(sample.keys())[0]])\n",
    "\n",
    "    # get max number of chunks for batch\n",
    "    if batch_total_length >= chunk_length:\n",
    "        batch_chunk_length = (batch_total_length // chunk_length) * chunk_length\n",
    "\n",
    "    # Split by chunks of max_len.\n",
    "    result = {\n",
    "        k: [t[i : i + chunk_length] for i in range(0, batch_chunk_length, chunk_length)]\n",
    "        for k, t in concatenated_examples.items()\n",
    "    }\n",
    "    # add remainder to global variable for next batch\n",
    "    remainder = {k: concatenated_examples[k][batch_chunk_length:] for k in concatenated_examples.keys()}\n",
    "    # prepare labels\n",
    "    result[\"labels\"] = result[\"input_ids\"].copy()\n",
    "    return result"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\n",
    "\n",
    "To train our model, we need to convert our inputs (text) to token IDs. This is done by a Hugging Face Transformers Tokenizer. In addition to QLoRA, we will use bitsanbytes 4-bit precision to quantize out frozen LLM to 4-bit and attach LoRA adapters on it.\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-11-15T09:26:20.833834Z",
     "start_time": "2023-11-15T09:26:20.830187Z"
    },
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "def print_trainable_parameters(model):\n",
    "    \"\"\"\n",
    "    Prints the number of trainable parameters in the model.\n",
    "    \"\"\"\n",
    "    trainable_params = 0\n",
    "    all_param = 0\n",
    "    for _, param in model.named_parameters():\n",
    "        all_param += param.numel()\n",
    "        if param.requires_grad:\n",
    "            trainable_params += param.numel()\n",
    "    print(\n",
    "        f\"trainable params: {trainable_params} || all params: {all_param} || trainable%: {100 * trainable_params / all_param}\"\n",
    "    )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Utility method for finding the target modules and update the necessary matrices. Visit [this](https://github.com/artidoro/qlora/blob/main/qlora.py) link for additional info."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-11-15T09:26:43.472318Z",
     "start_time": "2023-11-15T09:26:40.383333Z"
    },
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "import bitsandbytes as bnb\n",
    "\n",
    "def find_all_linear_names(hf_model):\n",
    "    lora_module_names = set()\n",
    "    for name, module in hf_model.named_modules():\n",
    "        if isinstance(module, bnb.nn.Linear4bit):\n",
    "            names = name.split(\".\")\n",
    "            lora_module_names.add(names[0] if len(names) == 1 else names[-1])\n",
    "\n",
    "    if \"lm_head\" in lora_module_names:  # needed for 16-bit\n",
    "        lora_module_names.remove(\"lm_head\")\n",
    "    return list(lora_module_names)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Define the train function"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-11-15T09:28:07.120756Z",
     "start_time": "2023-11-15T09:28:04.184835Z"
    },
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "from accelerate import Accelerator\n",
    "from huggingface_hub import login\n",
    "from peft import AutoPeftModelForCausalLM, LoraConfig, get_peft_model, prepare_model_for_kbit_training\n",
    "from sagemaker.remote_function import remote\n",
    "import torch\n",
    "from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig, set_seed\n",
    "import transformers\n",
    "\n",
    "# Start training\n",
    "@remote(volume_size=100, job_name_prefix=f\"train-{model_id.split('/')[-1].replace('.', '-')}\")\n",
    "def train_fn(\n",
    "        model_name,\n",
    "        train_ds,\n",
    "        test_ds=None,\n",
    "        lora_r=8,\n",
    "        lora_alpha=16,\n",
    "        lora_dropout=0.1,\n",
    "        per_device_train_batch_size=8,\n",
    "        per_device_eval_batch_size=8,\n",
    "        gradient_accumulation_steps=1,\n",
    "        learning_rate=2e-4,\n",
    "        num_train_epochs=1,\n",
    "        chunk_size=2048,\n",
    "        gradient_checkpointing=False,\n",
    "        merge_weights=False,\n",
    "        seed=42,\n",
    "        token=None\n",
    "):\n",
    "    set_seed(seed)\n",
    "\n",
    "    accelerator = Accelerator()\n",
    "\n",
    "    if token is not None:\n",
    "        login(token=token)\n",
    "\n",
    "    tokenizer = AutoTokenizer.from_pretrained(model_name)\n",
    "\n",
    "    # Set Tokenizer pad Token\n",
    "    tokenizer.pad_token = tokenizer.eos_token\n",
    "\n",
    "    with accelerator.main_process_first():\n",
    "        # tokenize and chunk dataset\n",
    "        lm_train_dataset = train_ds.map(\n",
    "            lambda sample: tokenizer(sample[\"text\"]), batched=True, remove_columns=list(train_ds.features)\n",
    "        ).map(\n",
    "            partial(chunk, chunk_length=chunk_size),\n",
    "            batched=True,\n",
    "        )\n",
    "    \n",
    "        if test_ds is not None:\n",
    "            lm_test_dataset = test_ds.map(\n",
    "                lambda sample: tokenizer(sample[\"text\"]), batched=True, remove_columns=list(test_ds.features)\n",
    "            ).map(\n",
    "                partial(chunk, chunk_length=chunk_size),\n",
    "                batched=True,\n",
    "            )\n",
    "    \n",
    "            print(f\"Total number of test samples: {len(lm_test_dataset)}\")\n",
    "        else:\n",
    "            lm_test_dataset = None\n",
    "\n",
    "    bnb_config = BitsAndBytesConfig(\n",
    "        load_in_4bit=True,\n",
    "        bnb_4bit_use_double_quant=True,\n",
    "        bnb_4bit_quant_type=\"nf4\",\n",
    "        bnb_4bit_compute_dtype=torch.bfloat16\n",
    "    )\n",
    "    \n",
    "    model = AutoModelForCausalLM.from_pretrained(\n",
    "        model_name,\n",
    "        trust_remote_code=True,\n",
    "        quantization_config=bnb_config,\n",
    "        attn_implementation=\"flash_attention_2\",\n",
    "        use_cache=False if gradient_checkpointing else True,\n",
    "        cache_dir=\"/tmp/.cache\"\n",
    "    )\n",
    "\n",
    "    model = prepare_model_for_kbit_training(model, use_gradient_checkpointing=gradient_checkpointing)\n",
    "\n",
    "    if gradient_checkpointing:\n",
    "        model.gradient_checkpointing_enable()\n",
    "\n",
    "    # get lora target modules\n",
    "    modules = find_all_linear_names(model)\n",
    "    print(f\"Found {len(modules)} modules to quantize: {modules}\")\n",
    "    \n",
    "    config = LoraConfig(\n",
    "        r=lora_r,\n",
    "        lora_alpha=lora_alpha,\n",
    "        target_modules=modules,\n",
    "        lora_dropout=lora_dropout,\n",
    "        bias=\"none\",\n",
    "        task_type=\"CAUSAL_LM\"\n",
    "    )\n",
    "\n",
    "    model = get_peft_model(model, config)\n",
    "    print_trainable_parameters(model)\n",
    "    \n",
    "    trainer = transformers.Trainer(\n",
    "        model=model,\n",
    "        train_dataset=lm_train_dataset,\n",
    "        eval_dataset=lm_test_dataset if lm_test_dataset is not None else None,\n",
    "        args=transformers.TrainingArguments(\n",
    "            per_device_train_batch_size=per_device_train_batch_size,\n",
    "            per_device_eval_batch_size=per_device_eval_batch_size,\n",
    "            gradient_accumulation_steps=gradient_accumulation_steps,\n",
    "            gradient_checkpointing=gradient_checkpointing,\n",
    "            logging_strategy=\"steps\",\n",
    "            logging_steps=1,\n",
    "            log_on_each_node=False,\n",
    "            num_train_epochs=num_train_epochs,\n",
    "            learning_rate=learning_rate,\n",
    "            bf16=True,\n",
    "            ddp_find_unused_parameters=False,\n",
    "            save_strategy=\"no\",\n",
    "            output_dir=\"outputs\"\n",
    "        ),\n",
    "        data_collator=transformers.DataCollatorForLanguageModeling(tokenizer, mlm=False),\n",
    "    )\n",
    "\n",
    "    trainer.train()\n",
    "\n",
    "    if merge_weights:\n",
    "        output_dir = \"/tmp/model\"\n",
    "\n",
    "        # merge adapter weights with base model and save\n",
    "        # save int 4 model\n",
    "        trainer.model.save_pretrained(output_dir, safe_serialization=False)\n",
    "        \n",
    "        if accelerator.is_main_process:\n",
    "            # clear memory\n",
    "            del model\n",
    "            del trainer\n",
    "    \n",
    "            torch.cuda.empty_cache()\n",
    "    \n",
    "            # load PEFT model\n",
    "            model = AutoPeftModelForCausalLM.from_pretrained(\n",
    "                output_dir,\n",
    "                torch_dtype=torch.float32,\n",
    "                low_cpu_mem_usage=True,\n",
    "                trust_remote_code=True,\n",
    "            )\n",
    "    \n",
    "            # Merge LoRA and base model and save\n",
    "            model = model.merge_and_unload()\n",
    "            model.save_pretrained(\n",
    "                \"/opt/ml/model\", safe_serialization=True, max_shard_size=\"2GB\"\n",
    "            )\n",
    "    else:\n",
    "        trainer.model.save_pretrained(\"/opt/ml/model\", safe_serialization=True)\n",
    "\n",
    "    if accelerator.is_main_process:\n",
    "        tokenizer.save_pretrained(\"/opt/ml/model\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-09-08T13:12:12.671800Z",
     "start_time": "2023-09-08T12:42:37.198731Z"
    },
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    },
    "scrolled": true,
    "tags": []
   },
   "outputs": [],
   "source": [
    "train_fn(\n",
    "    model_id,\n",
    "    train_ds=train_dataset,\n",
    "    per_device_train_batch_size=2,\n",
    "    per_device_eval_batch_size=1,\n",
    "    gradient_accumulation_steps=2,\n",
    "    gradient_checkpointing=True,\n",
    "    chunk_size=2048,\n",
    "    num_train_epochs=3,\n",
    "    merge_weights=True,\n",
    "    token=\"<HF_TOKEN>\"\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "tags": []
   },
   "source": [
    "## Load Fine-Tuned model\n",
    "\n",
    "Note: Run `train_fn` with `merge_weights=False`"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Download model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-09-11T14:39:39.945700Z",
     "start_time": "2023-09-11T14:39:39.726434Z"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "import boto3\n",
    "\n",
    "s3_client = boto3.client(\"s3\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-09-08T13:26:17.499800Z",
     "start_time": "2023-09-08T13:26:17.492889Z"
    },
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "model_id = \"meta-llama/Llama-2-13b-chat-hf\"\n",
    "\n",
    "bucket_name = \"<S3_BUCKET>\"\n",
    "job_prefix = f\"train-{model_id.split('/')[-1].replace('.', '-')}\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_last_job_name(job_name_prefix):\n",
    "    import boto3\n",
    "    sagemaker_client = boto3.client('sagemaker')\n",
    "    \n",
    "    search_response = sagemaker_client.search(\n",
    "        Resource='TrainingJob',\n",
    "        SearchExpression={\n",
    "            'Filters': [\n",
    "                {\n",
    "                    'Name': 'TrainingJobName',\n",
    "                    'Operator': 'Contains',\n",
    "                    'Value': job_name_prefix\n",
    "                },\n",
    "                {\n",
    "                    'Name': 'TrainingJobStatus',\n",
    "                    'Operator': 'Equals',\n",
    "                    'Value': \"Completed\"\n",
    "                }\n",
    "            ]\n",
    "        },\n",
    "        SortBy='CreationTime',\n",
    "        SortOrder='Descending',\n",
    "        MaxResults=1)\n",
    "    \n",
    "    return search_response['Results'][0]['TrainingJob']['TrainingJobName']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "job_name = get_last_job_name(job_prefix)\n",
    "\n",
    "job_name"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Donwload fine-tuned Peft model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-09-08T13:27:42.993432Z",
     "start_time": "2023-09-08T13:26:18.536625Z"
    },
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "s3_client.download_file(bucket_name, f\"{job_name}/{job_name}/output/model.tar.gz\", \"model.tar.gz\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-09-08T13:41:52.420226Z",
     "start_time": "2023-09-08T13:41:47.968834Z"
    },
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "! rm -rf ./model && mkdir -p ./model && tar -xf model.tar.gz -C ./model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we are loading the PEFT weights trained"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-09-11T14:39:44.140968Z",
     "start_time": "2023-09-11T14:39:43.661599Z"
    },
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "! huggingface-cli login --token <HF_TOKEN>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-09-11T14:39:45.562305Z",
     "start_time": "2023-09-11T14:39:44.594253Z"
    },
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "from transformers import AutoTokenizer\n",
    "\n",
    "model_id = \"meta-llama/Llama-2-13b-chat-hf\"\n",
    "\n",
    "tokenizer = AutoTokenizer.from_pretrained(model_id)\n",
    "\n",
    "tokenizer.pad_token = tokenizer.eos_token"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-07-20T20:16:01.359687Z",
     "start_time": "2023-07-20T20:09:58.562422Z"
    },
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "from peft import PeftModel, PeftConfig\n",
    "import torch\n",
    "from transformers import AutoModelForCausalLM\n",
    "\n",
    "device = 'cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu'\n",
    "\n",
    "config = PeftConfig.from_pretrained(\"./model\")\n",
    "model = AutoModelForCausalLM.from_pretrained(config.base_model_name_or_path, trust_remote_code=True)\n",
    "model = PeftModel.from_pretrained(model, \"./model\")\n",
    "model.to(device)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Load a test dataset and try a random sample for Q&A."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-07-20T20:16:01.885446Z",
     "start_time": "2023-07-20T20:16:01.323572Z"
    },
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# format sample\n",
    "prompt = f\"\"\"\n",
    "<s>[INST] {{question}} [/INST]\n",
    "\"\"\"\n",
    "\n",
    "test_sample = prompt.format(question=\"What are Amazon Bedrock Agents?\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-07-20T20:16:01.922060Z",
     "start_time": "2023-07-20T20:16:01.889875Z"
    },
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "input_ids = tokenizer(test_sample, return_tensors=\"pt\").input_ids"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "#set the tokens for the summary evaluation\n",
    "tokens_for_answer = 150\n",
    "output_tokens = input_ids.shape[1] + tokens_for_answer\n",
    "\n",
    "outputs = model.generate(inputs=input_ids.to(device), do_sample=True, max_length=output_tokens)\n",
    "gen_text = tokenizer.batch_decode(outputs)[0]\n",
    "\n",
    "print(gen_text)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    }
   },
   "source": [
    "## Deploy Fine-Tuned model\n",
    "\n",
    "Note: Run `train_fn` with `merge_weights=True`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-11-15T09:53:19.742445Z",
     "start_time": "2023-11-15T09:53:19.729286Z"
    },
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    }
   },
   "outputs": [],
   "source": [
    "import json\n",
    "import sagemaker\n",
    "from sagemaker import get_execution_role\n",
    "from sagemaker.huggingface import HuggingFaceModel, get_huggingface_llm_image_uri"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "sagemaker_session = sagemaker.Session()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-11-15T09:53:20.385178Z",
     "start_time": "2023-11-15T09:53:20.374889Z"
    },
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    }
   },
   "outputs": [],
   "source": [
    "model_id = \"meta-llama/Llama-2-13b-chat-hf\"\n",
    "\n",
    "bucket_name = sagemaker_session.default_bucket()\n",
    "job_prefix = f\"train-{model_id.split('/')[-1].replace('.', '-')}-merge\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_last_job_name(job_name_prefix):\n",
    "    import boto3\n",
    "    sagemaker_client = boto3.client('sagemaker')\n",
    "    \n",
    "    search_response = sagemaker_client.search(\n",
    "        Resource='TrainingJob',\n",
    "        SearchExpression={\n",
    "            'Filters': [\n",
    "                {\n",
    "                    'Name': 'TrainingJobName',\n",
    "                    'Operator': 'Contains',\n",
    "                    'Value': job_name_prefix\n",
    "                },\n",
    "                {\n",
    "                    'Name': 'TrainingJobStatus',\n",
    "                    'Operator': 'Equals',\n",
    "                    'Value': \"Completed\"\n",
    "                }\n",
    "            ]\n",
    "        },\n",
    "        SortBy='CreationTime',\n",
    "        SortOrder='Descending',\n",
    "        MaxResults=1)\n",
    "    \n",
    "    return search_response['Results'][0]['TrainingJob']['TrainingJobName']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "job_name = get_last_job_name(job_prefix)\n",
    "\n",
    "job_name"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    }
   },
   "source": [
    "### Inference configurations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-11-15T09:53:58.823709Z",
     "start_time": "2023-11-15T09:53:58.814508Z"
    },
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    }
   },
   "outputs": [],
   "source": [
    "instance_count = 1\n",
    "instance_type = \"ml.g5.48xlarge\"\n",
    "number_of_gpu = 8\n",
    "health_check_timeout = 300"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-11-15T09:54:00.992352Z",
     "start_time": "2023-11-15T09:54:00.673636Z"
    },
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    }
   },
   "outputs": [],
   "source": [
    "image_uri = get_huggingface_llm_image_uri(\n",
    "    \"huggingface\",\n",
    "    version=\"1.4\"\n",
    ")\n",
    "\n",
    "image_uri"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-11-15T09:54:05.067618Z",
     "start_time": "2023-11-15T09:54:04.429164Z"
    },
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    }
   },
   "outputs": [],
   "source": [
    "model = HuggingFaceModel(\n",
    "    image_uri=image_uri,\n",
    "    model_data=f\"s3://{bucket_name}/{job_name}/{job_name}/output/model.tar.gz\",\n",
    "    env={\n",
    "        'HF_MODEL_ID': \"/opt/ml/model\", # path to where sagemaker stores the model\n",
    "        'SM_NUM_GPUS': json.dumps(number_of_gpu), # Number of GPU used per replica\n",
    "    }\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-11-15T10:00:41.516254Z",
     "start_time": "2023-11-15T09:54:06.645438Z"
    },
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    }
   },
   "outputs": [],
   "source": [
    "predictor = model.deploy(\n",
    "    initial_instance_count=instance_count,\n",
    "    instance_type=instance_type,\n",
    "    container_startup_health_check_timeout=health_check_timeout,\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    }
   },
   "source": [
    "### Predict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-11-15T10:14:52.444752Z",
     "start_time": "2023-11-15T10:14:52.438786Z"
    },
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    }
   },
   "outputs": [],
   "source": [
    "base_prompt = f\"\"\"\n",
    "<s>[INST]\n",
    "{{question}} \n",
    "[/INST]\n",
    "\"\"\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-11-15T10:14:59.355976Z",
     "start_time": "2023-11-15T10:14:57.529876Z"
    },
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    }
   },
   "outputs": [],
   "source": [
    "prompt = base_prompt.format(question=\"What are Amazon Bedrock Agents?\")\n",
    "\n",
    "predictor.predict({\n",
    "\t\"inputs\": prompt,\n",
    "    \"parameters\": {\n",
    "        \"max_new_tokens\": 2048 - len(prompt),\n",
    "        \"temperature\": 0.2,\n",
    "        \"top_p\": 0.9,\n",
    "        \"stop\": [\".\"]\n",
    "    }\n",
    "})"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    }
   },
   "source": [
    "#### Delete Endpoint"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-11-15T10:15:18.381111Z",
     "start_time": "2023-11-15T10:15:17.187067Z"
    },
    "collapsed": false,
    "jupyter": {
     "outputs_hidden": false
    }
   },
   "outputs": [],
   "source": [
    "predictor.delete_model()\n",
    "predictor.delete_endpoint(delete_endpoint_config=True)"
   ]
  }
 ],
 "metadata": {
  "availableInstances": [
   {
    "_defaultOrder": 0,
    "_isFastLaunch": true,
    "category": "General purpose",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 4,
    "name": "ml.t3.medium",
    "vcpuNum": 2
   },
   {
    "_defaultOrder": 1,
    "_isFastLaunch": false,
    "category": "General purpose",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 8,
    "name": "ml.t3.large",
    "vcpuNum": 2
   },
   {
    "_defaultOrder": 2,
    "_isFastLaunch": false,
    "category": "General purpose",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 16,
    "name": "ml.t3.xlarge",
    "vcpuNum": 4
   },
   {
    "_defaultOrder": 3,
    "_isFastLaunch": false,
    "category": "General purpose",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 32,
    "name": "ml.t3.2xlarge",
    "vcpuNum": 8
   },
   {
    "_defaultOrder": 4,
    "_isFastLaunch": true,
    "category": "General purpose",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 8,
    "name": "ml.m5.large",
    "vcpuNum": 2
   },
   {
    "_defaultOrder": 5,
    "_isFastLaunch": false,
    "category": "General purpose",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 16,
    "name": "ml.m5.xlarge",
    "vcpuNum": 4
   },
   {
    "_defaultOrder": 6,
    "_isFastLaunch": false,
    "category": "General purpose",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 32,
    "name": "ml.m5.2xlarge",
    "vcpuNum": 8
   },
   {
    "_defaultOrder": 7,
    "_isFastLaunch": false,
    "category": "General purpose",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 64,
    "name": "ml.m5.4xlarge",
    "vcpuNum": 16
   },
   {
    "_defaultOrder": 8,
    "_isFastLaunch": false,
    "category": "General purpose",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 128,
    "name": "ml.m5.8xlarge",
    "vcpuNum": 32
   },
   {
    "_defaultOrder": 9,
    "_isFastLaunch": false,
    "category": "General purpose",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 192,
    "name": "ml.m5.12xlarge",
    "vcpuNum": 48
   },
   {
    "_defaultOrder": 10,
    "_isFastLaunch": false,
    "category": "General purpose",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 256,
    "name": "ml.m5.16xlarge",
    "vcpuNum": 64
   },
   {
    "_defaultOrder": 11,
    "_isFastLaunch": false,
    "category": "General purpose",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 384,
    "name": "ml.m5.24xlarge",
    "vcpuNum": 96
   },
   {
    "_defaultOrder": 12,
    "_isFastLaunch": false,
    "category": "General purpose",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 8,
    "name": "ml.m5d.large",
    "vcpuNum": 2
   },
   {
    "_defaultOrder": 13,
    "_isFastLaunch": false,
    "category": "General purpose",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 16,
    "name": "ml.m5d.xlarge",
    "vcpuNum": 4
   },
   {
    "_defaultOrder": 14,
    "_isFastLaunch": false,
    "category": "General purpose",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 32,
    "name": "ml.m5d.2xlarge",
    "vcpuNum": 8
   },
   {
    "_defaultOrder": 15,
    "_isFastLaunch": false,
    "category": "General purpose",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 64,
    "name": "ml.m5d.4xlarge",
    "vcpuNum": 16
   },
   {
    "_defaultOrder": 16,
    "_isFastLaunch": false,
    "category": "General purpose",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 128,
    "name": "ml.m5d.8xlarge",
    "vcpuNum": 32
   },
   {
    "_defaultOrder": 17,
    "_isFastLaunch": false,
    "category": "General purpose",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 192,
    "name": "ml.m5d.12xlarge",
    "vcpuNum": 48
   },
   {
    "_defaultOrder": 18,
    "_isFastLaunch": false,
    "category": "General purpose",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 256,
    "name": "ml.m5d.16xlarge",
    "vcpuNum": 64
   },
   {
    "_defaultOrder": 19,
    "_isFastLaunch": false,
    "category": "General purpose",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 384,
    "name": "ml.m5d.24xlarge",
    "vcpuNum": 96
   },
   {
    "_defaultOrder": 20,
    "_isFastLaunch": false,
    "category": "General purpose",
    "gpuNum": 0,
    "hideHardwareSpecs": true,
    "memoryGiB": 0,
    "name": "ml.geospatial.interactive",
    "supportedImageNames": [
     "sagemaker-geospatial-v1-0"
    ],
    "vcpuNum": 0
   },
   {
    "_defaultOrder": 21,
    "_isFastLaunch": true,
    "category": "Compute optimized",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 4,
    "name": "ml.c5.large",
    "vcpuNum": 2
   },
   {
    "_defaultOrder": 22,
    "_isFastLaunch": false,
    "category": "Compute optimized",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 8,
    "name": "ml.c5.xlarge",
    "vcpuNum": 4
   },
   {
    "_defaultOrder": 23,
    "_isFastLaunch": false,
    "category": "Compute optimized",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 16,
    "name": "ml.c5.2xlarge",
    "vcpuNum": 8
   },
   {
    "_defaultOrder": 24,
    "_isFastLaunch": false,
    "category": "Compute optimized",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 32,
    "name": "ml.c5.4xlarge",
    "vcpuNum": 16
   },
   {
    "_defaultOrder": 25,
    "_isFastLaunch": false,
    "category": "Compute optimized",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 72,
    "name": "ml.c5.9xlarge",
    "vcpuNum": 36
   },
   {
    "_defaultOrder": 26,
    "_isFastLaunch": false,
    "category": "Compute optimized",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 96,
    "name": "ml.c5.12xlarge",
    "vcpuNum": 48
   },
   {
    "_defaultOrder": 27,
    "_isFastLaunch": false,
    "category": "Compute optimized",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 144,
    "name": "ml.c5.18xlarge",
    "vcpuNum": 72
   },
   {
    "_defaultOrder": 28,
    "_isFastLaunch": false,
    "category": "Compute optimized",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 192,
    "name": "ml.c5.24xlarge",
    "vcpuNum": 96
   },
   {
    "_defaultOrder": 29,
    "_isFastLaunch": true,
    "category": "Accelerated computing",
    "gpuNum": 1,
    "hideHardwareSpecs": false,
    "memoryGiB": 16,
    "name": "ml.g4dn.xlarge",
    "vcpuNum": 4
   },
   {
    "_defaultOrder": 30,
    "_isFastLaunch": false,
    "category": "Accelerated computing",
    "gpuNum": 1,
    "hideHardwareSpecs": false,
    "memoryGiB": 32,
    "name": "ml.g4dn.2xlarge",
    "vcpuNum": 8
   },
   {
    "_defaultOrder": 31,
    "_isFastLaunch": false,
    "category": "Accelerated computing",
    "gpuNum": 1,
    "hideHardwareSpecs": false,
    "memoryGiB": 64,
    "name": "ml.g4dn.4xlarge",
    "vcpuNum": 16
   },
   {
    "_defaultOrder": 32,
    "_isFastLaunch": false,
    "category": "Accelerated computing",
    "gpuNum": 1,
    "hideHardwareSpecs": false,
    "memoryGiB": 128,
    "name": "ml.g4dn.8xlarge",
    "vcpuNum": 32
   },
   {
    "_defaultOrder": 33,
    "_isFastLaunch": false,
    "category": "Accelerated computing",
    "gpuNum": 4,
    "hideHardwareSpecs": false,
    "memoryGiB": 192,
    "name": "ml.g4dn.12xlarge",
    "vcpuNum": 48
   },
   {
    "_defaultOrder": 34,
    "_isFastLaunch": false,
    "category": "Accelerated computing",
    "gpuNum": 1,
    "hideHardwareSpecs": false,
    "memoryGiB": 256,
    "name": "ml.g4dn.16xlarge",
    "vcpuNum": 64
   },
   {
    "_defaultOrder": 35,
    "_isFastLaunch": false,
    "category": "Accelerated computing",
    "gpuNum": 1,
    "hideHardwareSpecs": false,
    "memoryGiB": 61,
    "name": "ml.p3.2xlarge",
    "vcpuNum": 8
   },
   {
    "_defaultOrder": 36,
    "_isFastLaunch": false,
    "category": "Accelerated computing",
    "gpuNum": 4,
    "hideHardwareSpecs": false,
    "memoryGiB": 244,
    "name": "ml.p3.8xlarge",
    "vcpuNum": 32
   },
   {
    "_defaultOrder": 37,
    "_isFastLaunch": false,
    "category": "Accelerated computing",
    "gpuNum": 8,
    "hideHardwareSpecs": false,
    "memoryGiB": 488,
    "name": "ml.p3.16xlarge",
    "vcpuNum": 64
   },
   {
    "_defaultOrder": 38,
    "_isFastLaunch": false,
    "category": "Accelerated computing",
    "gpuNum": 8,
    "hideHardwareSpecs": false,
    "memoryGiB": 768,
    "name": "ml.p3dn.24xlarge",
    "vcpuNum": 96
   },
   {
    "_defaultOrder": 39,
    "_isFastLaunch": false,
    "category": "Memory Optimized",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 16,
    "name": "ml.r5.large",
    "vcpuNum": 2
   },
   {
    "_defaultOrder": 40,
    "_isFastLaunch": false,
    "category": "Memory Optimized",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 32,
    "name": "ml.r5.xlarge",
    "vcpuNum": 4
   },
   {
    "_defaultOrder": 41,
    "_isFastLaunch": false,
    "category": "Memory Optimized",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 64,
    "name": "ml.r5.2xlarge",
    "vcpuNum": 8
   },
   {
    "_defaultOrder": 42,
    "_isFastLaunch": false,
    "category": "Memory Optimized",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 128,
    "name": "ml.r5.4xlarge",
    "vcpuNum": 16
   },
   {
    "_defaultOrder": 43,
    "_isFastLaunch": false,
    "category": "Memory Optimized",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 256,
    "name": "ml.r5.8xlarge",
    "vcpuNum": 32
   },
   {
    "_defaultOrder": 44,
    "_isFastLaunch": false,
    "category": "Memory Optimized",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 384,
    "name": "ml.r5.12xlarge",
    "vcpuNum": 48
   },
   {
    "_defaultOrder": 45,
    "_isFastLaunch": false,
    "category": "Memory Optimized",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 512,
    "name": "ml.r5.16xlarge",
    "vcpuNum": 64
   },
   {
    "_defaultOrder": 46,
    "_isFastLaunch": false,
    "category": "Memory Optimized",
    "gpuNum": 0,
    "hideHardwareSpecs": false,
    "memoryGiB": 768,
    "name": "ml.r5.24xlarge",
    "vcpuNum": 96
   },
   {
    "_defaultOrder": 47,
    "_isFastLaunch": false,
    "category": "Accelerated computing",
    "gpuNum": 1,
    "hideHardwareSpecs": false,
    "memoryGiB": 16,
    "name": "ml.g5.xlarge",
    "vcpuNum": 4
   },
   {
    "_defaultOrder": 48,
    "_isFastLaunch": false,
    "category": "Accelerated computing",
    "gpuNum": 1,
    "hideHardwareSpecs": false,
    "memoryGiB": 32,
    "name": "ml.g5.2xlarge",
    "vcpuNum": 8
   },
   {
    "_defaultOrder": 49,
    "_isFastLaunch": false,
    "category": "Accelerated computing",
    "gpuNum": 1,
    "hideHardwareSpecs": false,
    "memoryGiB": 64,
    "name": "ml.g5.4xlarge",
    "vcpuNum": 16
   },
   {
    "_defaultOrder": 50,
    "_isFastLaunch": false,
    "category": "Accelerated computing",
    "gpuNum": 1,
    "hideHardwareSpecs": false,
    "memoryGiB": 128,
    "name": "ml.g5.8xlarge",
    "vcpuNum": 32
   },
   {
    "_defaultOrder": 51,
    "_isFastLaunch": false,
    "category": "Accelerated computing",
    "gpuNum": 1,
    "hideHardwareSpecs": false,
    "memoryGiB": 256,
    "name": "ml.g5.16xlarge",
    "vcpuNum": 64
   },
   {
    "_defaultOrder": 52,
    "_isFastLaunch": false,
    "category": "Accelerated computing",
    "gpuNum": 4,
    "hideHardwareSpecs": false,
    "memoryGiB": 192,
    "name": "ml.g5.12xlarge",
    "vcpuNum": 48
   },
   {
    "_defaultOrder": 53,
    "_isFastLaunch": false,
    "category": "Accelerated computing",
    "gpuNum": 4,
    "hideHardwareSpecs": false,
    "memoryGiB": 384,
    "name": "ml.g5.24xlarge",
    "vcpuNum": 96
   },
   {
    "_defaultOrder": 54,
    "_isFastLaunch": false,
    "category": "Accelerated computing",
    "gpuNum": 8,
    "hideHardwareSpecs": false,
    "memoryGiB": 768,
    "name": "ml.g5.48xlarge",
    "vcpuNum": 192
   },
   {
    "_defaultOrder": 55,
    "_isFastLaunch": false,
    "category": "Accelerated computing",
    "gpuNum": 8,
    "hideHardwareSpecs": false,
    "memoryGiB": 1152,
    "name": "ml.p4d.24xlarge",
    "vcpuNum": 96
   },
   {
    "_defaultOrder": 56,
    "_isFastLaunch": false,
    "category": "Accelerated computing",
    "gpuNum": 8,
    "hideHardwareSpecs": false,
    "memoryGiB": 1152,
    "name": "ml.p4de.24xlarge",
    "vcpuNum": 96
   }
  ],
  "instance_type": "ml.g5.12xlarge",
  "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": "3.11.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}