Model_FineTuning.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Step 3: Model Fine-tuning\n",
    "In this notebook, you'll fine-tune the Meta Llama 2 7B large language model, deploy the fine-tuned model, and test it's text generation and domain knowledge capabilities. \n",
    "\n",
    "Fine-tuning refers to the process of taking a pre-trained language model and retraining it for a different but related task using specific data. This approach is also known as transfer learning, which involves transferring the knowledge learned from one task to another. Large language models (LLMs) like Llama 2 7B are trained on massive amounts of unlabeled data and can be fine-tuned on domain domain datasets, making the model perform better on that specific domain.\n",
    "\n",
    "Input: A train and an optional validation directory. Each directory contains a CSV/JSON/TXT file.\n",
    "For CSV/JSON files, the train or validation data is used from the column called 'text' or the first column if no column called 'text' is found.\n",
    "The number of files under train and validation should equal to one.\n",
    "\n",
    "- **You'll choose your dataset below based on the domain you've chosen**\n",
    "\n",
    "Output: A trained model that can be deployed for inference.\\\n",
    "After you've fine-tuned the model, you'll evaluate it with the same input you used in project step 2: model evaluation. \n",
    "\n",
    "---"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Set up\n",
    "\n",
    "---\n",
    "Install and import the necessary packages. Restart the kernel after executing the cell below. \n",
    "\n",
    "---"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: sagemaker in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (2.229.0)\n",
      "Requirement already satisfied: datasets in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (2.21.0)\n",
      "Requirement already satisfied: attrs<24,>=23.1.0 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from sagemaker) (23.2.0)\n",
      "Requirement already satisfied: boto3<2.0,>=1.34.142 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from sagemaker) (1.34.142)\n",
      "Requirement already satisfied: cloudpickle==2.2.1 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from sagemaker) (2.2.1)\n",
      "Requirement already satisfied: docker in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from sagemaker) (7.1.0)\n",
      "Requirement already satisfied: google-pasta in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from sagemaker) (0.2.0)\n",
      "Requirement already satisfied: importlib-metadata<7.0,>=1.4.0 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from sagemaker) (6.11.0)\n",
      "Requirement already satisfied: jsonschema in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from sagemaker) (4.22.0)\n",
      "Requirement already satisfied: numpy<2.0,>=1.9.0 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from sagemaker) (1.26.4)\n",
      "Requirement already satisfied: packaging>=20.0 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from sagemaker) (24.1)\n",
      "Requirement already satisfied: pandas in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from sagemaker) (1.5.3)\n",
      "Requirement already satisfied: pathos in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from sagemaker) (0.3.2)\n",
      "Requirement already satisfied: platformdirs in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from sagemaker) (4.2.2)\n",
      "Requirement already satisfied: protobuf<5.0,>=3.12 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from sagemaker) (4.25.3)\n",
      "Requirement already satisfied: psutil in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from sagemaker) (5.9.8)\n",
      "Requirement already satisfied: PyYAML~=6.0 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from sagemaker) (6.0.1)\n",
      "Requirement already satisfied: requests in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from sagemaker) (2.32.3)\n",
      "Requirement already satisfied: schema in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from sagemaker) (0.7.7)\n",
      "Requirement already satisfied: smdebug-rulesconfig==1.0.1 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from sagemaker) (1.0.1)\n",
      "Requirement already satisfied: tblib<4,>=1.7.0 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from sagemaker) (3.0.0)\n",
      "Requirement already satisfied: tqdm in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from sagemaker) (4.66.4)\n",
      "Requirement already satisfied: urllib3<3.0.0,>=1.26.8 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from sagemaker) (2.2.1)\n",
      "Requirement already satisfied: filelock in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from datasets) (3.14.0)\n",
      "Requirement already satisfied: pyarrow>=15.0.0 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from datasets) (16.1.0)\n",
      "Requirement already satisfied: dill<0.3.9,>=0.3.0 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from datasets) (0.3.8)\n",
      "Requirement already satisfied: xxhash in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from datasets) (3.5.0)\n",
      "Requirement already satisfied: multiprocess in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from datasets) (0.70.16)\n",
      "Requirement already satisfied: fsspec<=2024.6.1,>=2023.1.0 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from fsspec[http]<=2024.6.1,>=2023.1.0->datasets) (2024.6.0)\n",
      "Requirement already satisfied: aiohttp in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from datasets) (3.9.5)\n",
      "Requirement already satisfied: huggingface-hub>=0.21.2 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from datasets) (0.24.6)\n",
      "Requirement already satisfied: botocore<1.35.0,>=1.34.142 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from boto3<2.0,>=1.34.142->sagemaker) (1.34.142)\n",
      "Requirement already satisfied: jmespath<2.0.0,>=0.7.1 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from boto3<2.0,>=1.34.142->sagemaker) (1.0.1)\n",
      "Requirement already satisfied: s3transfer<0.11.0,>=0.10.0 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from boto3<2.0,>=1.34.142->sagemaker) (0.10.1)\n",
      "Requirement already satisfied: aiosignal>=1.1.2 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from aiohttp->datasets) (1.3.1)\n",
      "Requirement already satisfied: frozenlist>=1.1.1 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from aiohttp->datasets) (1.4.1)\n",
      "Requirement already satisfied: multidict<7.0,>=4.5 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from aiohttp->datasets) (6.0.5)\n",
      "Requirement already satisfied: yarl<2.0,>=1.0 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from aiohttp->datasets) (1.9.4)\n",
      "Requirement already satisfied: async-timeout<5.0,>=4.0 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from aiohttp->datasets) (4.0.3)\n",
      "Requirement already satisfied: typing-extensions>=3.7.4.3 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from huggingface-hub>=0.21.2->datasets) (4.12.1)\n",
      "Requirement already satisfied: zipp>=0.5 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from importlib-metadata<7.0,>=1.4.0->sagemaker) (3.17.0)\n",
      "Requirement already satisfied: charset-normalizer<4,>=2 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from requests->sagemaker) (3.3.2)\n",
      "Requirement already satisfied: idna<4,>=2.5 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from requests->sagemaker) (3.7)\n",
      "Requirement already satisfied: certifi>=2017.4.17 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from requests->sagemaker) (2024.2.2)\n",
      "Requirement already satisfied: six in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from google-pasta->sagemaker) (1.16.0)\n",
      "Requirement already satisfied: jsonschema-specifications>=2023.03.6 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from jsonschema->sagemaker) (2023.12.1)\n",
      "Requirement already satisfied: referencing>=0.28.4 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from jsonschema->sagemaker) (0.35.1)\n",
      "Requirement already satisfied: rpds-py>=0.7.1 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from jsonschema->sagemaker) (0.18.1)\n",
      "Requirement already satisfied: python-dateutil>=2.8.1 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from pandas->sagemaker) (2.9.0)\n",
      "Requirement already satisfied: pytz>=2020.1 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from pandas->sagemaker) (2024.1)\n",
      "Requirement already satisfied: ppft>=1.7.6.8 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from pathos->sagemaker) (1.7.6.8)\n",
      "Requirement already satisfied: pox>=0.3.4 in /home/ec2-user/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages (from pathos->sagemaker) (0.3.4)\n"
     ]
    }
   ],
   "source": [
    "!pip install --upgrade sagemaker datasets"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Select the model to fine-tune"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "model_id, model_version = \"meta-textgeneration-llama-2-7b\", \"2.*\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In the cell below, choose the training dataset text for the domain you've chosen and update the code in the cell below:  \n",
    "\n",
    "To create a finance domain expert model: \n",
    "\n",
    "- `\"training\": f\"s3://genaiwithawsproject2024/training-datasets/finance\"`\n",
    "\n",
    "To create a medical domain expert model: \n",
    "\n",
    "- `\"training\": f\"s3://genaiwithawsproject2024/training-datasets/medical\"`\n",
    "\n",
    "To create an IT domain expert model: \n",
    "\n",
    "- `\"training\": f\"s3://genaiwithawsproject2024/training-datasets/it\"`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sagemaker.config INFO - Not applying SDK defaults from location: /etc/xdg/sagemaker/config.yaml\n",
      "sagemaker.config INFO - Not applying SDK defaults from location: /home/ec2-user/.config/sagemaker/config.yaml\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using model 'meta-textgeneration-llama-2-7b' with wildcard version identifier '*'. You can pin to version '4.5.0' for more stable results. Note that models may have different input/output signatures after a major version upgrade.\n",
      "INFO:sagemaker:Creating training-job with name: meta-textgeneration-llama-2-7b-2024-08-20-07-28-32-383\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2024-08-20 07:28:38 Starting - Starting the training job\n",
      "2024-08-20 07:28:38 Pending - Training job waiting for capacity...\n",
      "2024-08-20 07:28:54 Pending - Preparing the instances for training...\n",
      "2024-08-20 07:29:19 Downloading - Downloading input data..............................\n",
      "2024-08-20 07:34:32 Training - Training image download completed. Training in progress..\u001b[34mbash: cannot set terminal process group (-1): Inappropriate ioctl for device\u001b[0m\n",
      "\u001b[34mbash: no job control in this shell\u001b[0m\n",
      "\u001b[34m2024-08-20 07:34:36,848 sagemaker-training-toolkit INFO     Imported framework sagemaker_pytorch_container.training\u001b[0m\n",
      "\u001b[34m2024-08-20 07:34:36,865 sagemaker-training-toolkit INFO     No Neurons detected (normal if no neurons installed)\u001b[0m\n",
      "\u001b[34m2024-08-20 07:34:36,875 sagemaker_pytorch_container.training INFO     Block until all host DNS lookups succeed.\u001b[0m\n",
      "\u001b[34m2024-08-20 07:34:36,879 sagemaker_pytorch_container.training INFO     Invoking user training script.\u001b[0m\n",
      "\u001b[34m2024-08-20 07:34:46,155 sagemaker-training-toolkit INFO     Installing dependencies from requirements.txt:\u001b[0m\n",
      "\u001b[34m/opt/conda/bin/python3.10 -m pip install -r requirements.txt\u001b[0m\n",
      "\u001b[34mProcessing ./lib/accelerate/accelerate-0.33.0-py3-none-any.whl (from -r requirements.txt (line 1))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/bitsandbytes/bitsandbytes-0.39.1-py3-none-any.whl (from -r requirements.txt (line 2))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/black/black-23.7.0-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (from -r requirements.txt (line 3))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/brotli/Brotli-1.0.9-cp310-cp310-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_12_x86_64.manylinux2010_x86_64.whl (from -r requirements.txt (line 4))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/datasets/datasets-2.14.1-py3-none-any.whl (from -r requirements.txt (line 5))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/docstring-parser/docstring_parser-0.16-py3-none-any.whl (from -r requirements.txt (line 6))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/fire/fire-0.5.0.tar.gz\u001b[0m\n",
      "\u001b[34mPreparing metadata (setup.py): started\u001b[0m\n",
      "\u001b[34mPreparing metadata (setup.py): finished with status 'done'\u001b[0m\n",
      "\u001b[34mProcessing ./lib/huggingface-hub/huggingface_hub-0.24.2-py3-none-any.whl (from -r requirements.txt (line 8))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/inflate64/inflate64-0.3.1-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (from -r requirements.txt (line 9))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/loralib/loralib-0.1.1-py3-none-any.whl (from -r requirements.txt (line 10))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/multivolumefile/multivolumefile-0.2.3-py3-none-any.whl (from -r requirements.txt (line 11))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/mypy-extensions/mypy_extensions-1.0.0-py3-none-any.whl (from -r requirements.txt (line 12))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/nvidia-cublas-cu12/nvidia_cublas_cu12-12.1.3.1-py3-none-manylinux1_x86_64.whl (from -r requirements.txt (line 13))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/nvidia-cuda-cupti-cu12/nvidia_cuda_cupti_cu12-12.1.105-py3-none-manylinux1_x86_64.whl (from -r requirements.txt (line 14))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/nvidia-cuda-nvrtc-cu12/nvidia_cuda_nvrtc_cu12-12.1.105-py3-none-manylinux1_x86_64.whl (from -r requirements.txt (line 15))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/nvidia-cuda-runtime-cu12/nvidia_cuda_runtime_cu12-12.1.105-py3-none-manylinux1_x86_64.whl (from -r requirements.txt (line 16))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/nvidia-cudnn-cu12/nvidia_cudnn_cu12-8.9.2.26-py3-none-manylinux1_x86_64.whl (from -r requirements.txt (line 17))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/nvidia-cufft-cu12/nvidia_cufft_cu12-11.0.2.54-py3-none-manylinux1_x86_64.whl (from -r requirements.txt (line 18))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/nvidia-curand-cu12/nvidia_curand_cu12-10.3.2.106-py3-none-manylinux1_x86_64.whl (from -r requirements.txt (line 19))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/nvidia-cusolver-cu12/nvidia_cusolver_cu12-11.4.5.107-py3-none-manylinux1_x86_64.whl (from -r requirements.txt (line 20))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/nvidia-cusparse-cu12/nvidia_cusparse_cu12-12.1.0.106-py3-none-manylinux1_x86_64.whl (from -r requirements.txt (line 21))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/nvidia-nccl-cu12/nvidia_nccl_cu12-2.19.3-py3-none-manylinux1_x86_64.whl (from -r requirements.txt (line 22))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/nvidia-nvjitlink-cu12/nvidia_nvjitlink_cu12-12.3.101-py3-none-manylinux1_x86_64.whl (from -r requirements.txt (line 23))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/nvidia-nvtx-cu12/nvidia_nvtx_cu12-12.1.105-py3-none-manylinux1_x86_64.whl (from -r requirements.txt (line 24))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/pathspec/pathspec-0.11.1-py3-none-any.whl (from -r requirements.txt (line 25))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/peft/peft-0.4.0-py3-none-any.whl (from -r requirements.txt (line 26))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/py7zr/py7zr-0.20.5-py3-none-any.whl (from -r requirements.txt (line 27))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/pybcj/pybcj-1.0.1-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (from -r requirements.txt (line 28))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/pycryptodomex/pycryptodomex-3.18.0-cp35-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (from -r requirements.txt (line 29))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/pyppmd/pyppmd-1.0.0-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (from -r requirements.txt (line 30))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/pyzstd/pyzstd-0.15.9-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (from -r requirements.txt (line 31))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/safetensors/safetensors-0.4.2-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (from -r requirements.txt (line 32))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/scipy/scipy-1.11.1-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (from -r requirements.txt (line 33))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/shtab/shtab-1.7.1-py3-none-any.whl (from -r requirements.txt (line 34))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/termcolor/termcolor-2.3.0-py3-none-any.whl (from -r requirements.txt (line 35))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/texttable/texttable-1.6.7-py2.py3-none-any.whl (from -r requirements.txt (line 36))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/tokenize-rt/tokenize_rt-5.1.0-py2.py3-none-any.whl (from -r requirements.txt (line 37))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/tokenizers/tokenizers-0.19.1-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (from -r requirements.txt (line 38))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/torch/torch-2.2.0-cp310-cp310-manylinux1_x86_64.whl (from -r requirements.txt (line 39))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/transformers/transformers-4.43.1-py3-none-any.whl (from -r requirements.txt (line 40))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/triton/triton-2.2.0-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (from -r requirements.txt (line 41))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/trl/trl-0.8.1-py3-none-any.whl (from -r requirements.txt (line 42))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/typing-extensions/typing_extensions-4.8.0-py3-none-any.whl (from -r requirements.txt (line 43))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/tyro/tyro-0.7.3-py3-none-any.whl (from -r requirements.txt (line 44))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/sagemaker_jumpstart_script_utilities/sagemaker_jumpstart_script_utilities-1.1.9-py2.py3-none-any.whl (from -r requirements.txt (line 45))\u001b[0m\n",
      "\u001b[34mProcessing ./lib/sagemaker_jumpstart_huggingface_script_utilities/sagemaker_jumpstart_huggingface_script_utilities-1.2.5-py2.py3-none-any.whl (from -r requirements.txt (line 46))\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: numpy<2.0.0,>=1.17 in /opt/conda/lib/python3.10/site-packages (from accelerate==0.33.0->-r requirements.txt (line 1)) (1.24.4)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: packaging>=20.0 in /opt/conda/lib/python3.10/site-packages (from accelerate==0.33.0->-r requirements.txt (line 1)) (23.1)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: psutil in /opt/conda/lib/python3.10/site-packages (from accelerate==0.33.0->-r requirements.txt (line 1)) (5.9.5)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: pyyaml in /opt/conda/lib/python3.10/site-packages (from accelerate==0.33.0->-r requirements.txt (line 1)) (6.0)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: click>=8.0.0 in /opt/conda/lib/python3.10/site-packages (from black==23.7.0->-r requirements.txt (line 3)) (8.1.4)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: platformdirs>=2 in /opt/conda/lib/python3.10/site-packages (from black==23.7.0->-r requirements.txt (line 3)) (3.8.1)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: tomli>=1.1.0 in /opt/conda/lib/python3.10/site-packages (from black==23.7.0->-r requirements.txt (line 3)) (2.0.1)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: pyarrow>=8.0.0 in /opt/conda/lib/python3.10/site-packages (from datasets==2.14.1->-r requirements.txt (line 5)) (14.0.2)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: dill<0.3.8,>=0.3.0 in /opt/conda/lib/python3.10/site-packages (from datasets==2.14.1->-r requirements.txt (line 5)) (0.3.6)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: pandas in /opt/conda/lib/python3.10/site-packages (from datasets==2.14.1->-r requirements.txt (line 5)) (2.0.3)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: requests>=2.19.0 in /opt/conda/lib/python3.10/site-packages (from datasets==2.14.1->-r requirements.txt (line 5)) (2.31.0)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: tqdm>=4.62.1 in /opt/conda/lib/python3.10/site-packages (from datasets==2.14.1->-r requirements.txt (line 5)) (4.65.0)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: xxhash in /opt/conda/lib/python3.10/site-packages (from datasets==2.14.1->-r requirements.txt (line 5)) (3.4.1)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: multiprocess in /opt/conda/lib/python3.10/site-packages (from datasets==2.14.1->-r requirements.txt (line 5)) (0.70.14)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: fsspec>=2021.11.1 in /opt/conda/lib/python3.10/site-packages (from fsspec[http]>=2021.11.1->datasets==2.14.1->-r requirements.txt (line 5)) (2023.6.0)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: aiohttp in /opt/conda/lib/python3.10/site-packages (from datasets==2.14.1->-r requirements.txt (line 5)) (3.9.3)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: six in /opt/conda/lib/python3.10/site-packages (from fire==0.5.0->-r requirements.txt (line 7)) (1.16.0)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: filelock in /opt/conda/lib/python3.10/site-packages (from huggingface-hub==0.24.2->-r requirements.txt (line 8)) (3.12.2)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: sympy in /opt/conda/lib/python3.10/site-packages (from torch==2.2.0->-r requirements.txt (line 39)) (1.12)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: networkx in /opt/conda/lib/python3.10/site-packages (from torch==2.2.0->-r requirements.txt (line 39)) (3.1)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: jinja2 in /opt/conda/lib/python3.10/site-packages (from torch==2.2.0->-r requirements.txt (line 39)) (3.1.2)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: regex!=2019.12.17 in /opt/conda/lib/python3.10/site-packages (from transformers==4.43.1->-r requirements.txt (line 40)) (2023.12.25)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: rich>=11.1.0 in /opt/conda/lib/python3.10/site-packages (from tyro==0.7.3->-r requirements.txt (line 44)) (13.4.2)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: aiosignal>=1.1.2 in /opt/conda/lib/python3.10/site-packages (from aiohttp->datasets==2.14.1->-r requirements.txt (line 5)) (1.3.1)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: attrs>=17.3.0 in /opt/conda/lib/python3.10/site-packages (from aiohttp->datasets==2.14.1->-r requirements.txt (line 5)) (23.1.0)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: frozenlist>=1.1.1 in /opt/conda/lib/python3.10/site-packages (from aiohttp->datasets==2.14.1->-r requirements.txt (line 5)) (1.4.1)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: multidict<7.0,>=4.5 in /opt/conda/lib/python3.10/site-packages (from aiohttp->datasets==2.14.1->-r requirements.txt (line 5)) (6.0.5)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: yarl<2.0,>=1.0 in /opt/conda/lib/python3.10/site-packages (from aiohttp->datasets==2.14.1->-r requirements.txt (line 5)) (1.9.4)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: async-timeout<5.0,>=4.0 in /opt/conda/lib/python3.10/site-packages (from aiohttp->datasets==2.14.1->-r requirements.txt (line 5)) (4.0.3)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: charset-normalizer<4,>=2 in /opt/conda/lib/python3.10/site-packages (from requests>=2.19.0->datasets==2.14.1->-r requirements.txt (line 5)) (3.1.0)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: idna<4,>=2.5 in /opt/conda/lib/python3.10/site-packages (from requests>=2.19.0->datasets==2.14.1->-r requirements.txt (line 5)) (3.4)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: urllib3<3,>=1.21.1 in /opt/conda/lib/python3.10/site-packages (from requests>=2.19.0->datasets==2.14.1->-r requirements.txt (line 5)) (1.26.15)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: certifi>=2017.4.17 in /opt/conda/lib/python3.10/site-packages (from requests>=2.19.0->datasets==2.14.1->-r requirements.txt (line 5)) (2024.2.2)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: markdown-it-py>=2.2.0 in /opt/conda/lib/python3.10/site-packages (from rich>=11.1.0->tyro==0.7.3->-r requirements.txt (line 44)) (3.0.0)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: pygments<3.0.0,>=2.13.0 in /opt/conda/lib/python3.10/site-packages (from rich>=11.1.0->tyro==0.7.3->-r requirements.txt (line 44)) (2.15.1)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: MarkupSafe>=2.0 in /opt/conda/lib/python3.10/site-packages (from jinja2->torch==2.2.0->-r requirements.txt (line 39)) (2.1.3)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: python-dateutil>=2.8.2 in /opt/conda/lib/python3.10/site-packages (from pandas->datasets==2.14.1->-r requirements.txt (line 5)) (2.8.2)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: pytz>=2020.1 in /opt/conda/lib/python3.10/site-packages (from pandas->datasets==2.14.1->-r requirements.txt (line 5)) (2023.3)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: tzdata>=2022.1 in /opt/conda/lib/python3.10/site-packages (from pandas->datasets==2.14.1->-r requirements.txt (line 5)) (2023.3)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: mpmath>=0.19 in /opt/conda/lib/python3.10/site-packages (from sympy->torch==2.2.0->-r requirements.txt (line 39)) (1.3.0)\u001b[0m\n",
      "\u001b[34mRequirement already satisfied: mdurl~=0.1 in /opt/conda/lib/python3.10/site-packages (from markdown-it-py>=2.2.0->rich>=11.1.0->tyro==0.7.3->-r requirements.txt (line 44)) (0.1.0)\u001b[0m\n",
      "\u001b[34mscipy is already installed with the same version as the provided wheel. Use --force-reinstall to force an installation of the wheel.\u001b[0m\n",
      "\u001b[34mBuilding wheels for collected packages: fire\u001b[0m\n",
      "\u001b[34mBuilding wheel for fire (setup.py): started\u001b[0m\n",
      "\u001b[34mBuilding wheel for fire (setup.py): finished with status 'done'\u001b[0m\n",
      "\u001b[34mCreated wheel for fire: filename=fire-0.5.0-py2.py3-none-any.whl size=116932 sha256=91761c9accb5971921180f45f35511f6b46101250d62242beec0dd545de61759\u001b[0m\n",
      "\u001b[34mStored in directory: /root/.cache/pip/wheels/db/3d/41/7e69dca5f61e37d109a4457082ffc5c6edb55ab633bafded38\u001b[0m\n",
      "\u001b[34mSuccessfully built fire\u001b[0m\n",
      "\u001b[34mInstalling collected packages: texttable, Brotli, bitsandbytes, typing-extensions, triton, tokenize-rt, termcolor, shtab, sagemaker-jumpstart-script-utilities, sagemaker-jumpstart-huggingface-script-utilities, safetensors, pyzstd, pyppmd, pycryptodomex, pybcj, pathspec, nvidia-nvtx-cu12, nvidia-nvjitlink-cu12, nvidia-nccl-cu12, nvidia-curand-cu12, nvidia-cufft-cu12, nvidia-cuda-runtime-cu12, nvidia-cuda-nvrtc-cu12, nvidia-cuda-cupti-cu12, nvidia-cublas-cu12, mypy-extensions, multivolumefile, loralib, inflate64, docstring-parser, py7zr, nvidia-cusparse-cu12, nvidia-cudnn-cu12, huggingface-hub, fire, black, tyro, tokenizers, nvidia-cusolver-cu12, transformers, torch, datasets, accelerate, trl, peft\u001b[0m\n",
      "\u001b[34mAttempting uninstall: typing-extensions\u001b[0m\n",
      "\u001b[34mFound existing installation: typing_extensions 4.7.1\u001b[0m\n",
      "\u001b[34mUninstalling typing_extensions-4.7.1:\u001b[0m\n",
      "\u001b[34mSuccessfully uninstalled typing_extensions-4.7.1\u001b[0m\n",
      "\u001b[34mAttempting uninstall: triton\u001b[0m\n",
      "\u001b[34mFound existing installation: triton 2.0.0.dev20221202\u001b[0m\n",
      "\u001b[34mUninstalling triton-2.0.0.dev20221202:\u001b[0m\n",
      "\u001b[34mSuccessfully uninstalled triton-2.0.0.dev20221202\u001b[0m\n",
      "\u001b[34mAttempting uninstall: huggingface-hub\u001b[0m\n",
      "\u001b[34mFound existing installation: huggingface-hub 0.20.3\u001b[0m\n",
      "\u001b[34mUninstalling huggingface-hub-0.20.3:\u001b[0m\n",
      "\u001b[34mSuccessfully uninstalled huggingface-hub-0.20.3\u001b[0m\n",
      "\u001b[34mAttempting uninstall: tokenizers\u001b[0m\n",
      "\u001b[34mFound existing installation: tokenizers 0.13.3\u001b[0m\n",
      "\u001b[34mUninstalling tokenizers-0.13.3:\u001b[0m\n",
      "\u001b[34mSuccessfully uninstalled tokenizers-0.13.3\u001b[0m\n",
      "\u001b[34mAttempting uninstall: transformers\u001b[0m\n",
      "\u001b[34mFound existing installation: transformers 4.28.1\u001b[0m\n",
      "\u001b[34mUninstalling transformers-4.28.1:\u001b[0m\n",
      "\u001b[34mSuccessfully uninstalled transformers-4.28.1\u001b[0m\n",
      "\u001b[34mAttempting uninstall: torch\u001b[0m\n",
      "\u001b[34mFound existing installation: torch 2.0.0\u001b[0m\n",
      "\u001b[34mUninstalling torch-2.0.0:\u001b[0m\n",
      "\u001b[34mSuccessfully uninstalled torch-2.0.0\u001b[0m\n",
      "\u001b[34mAttempting uninstall: datasets\u001b[0m\n",
      "\u001b[34mFound existing installation: datasets 2.16.1\u001b[0m\n",
      "\u001b[34mUninstalling datasets-2.16.1:\u001b[0m\n",
      "\u001b[34mSuccessfully uninstalled datasets-2.16.1\u001b[0m\n",
      "\u001b[34mAttempting uninstall: accelerate\u001b[0m\n",
      "\u001b[34mFound existing installation: accelerate 0.19.0\u001b[0m\n",
      "\u001b[34mUninstalling accelerate-0.19.0:\u001b[0m\n",
      "\u001b[34mSuccessfully uninstalled accelerate-0.19.0\u001b[0m\n",
      "\u001b[34mERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.\u001b[0m\n",
      "\u001b[34mfastai 2.7.12 requires torch<2.1,>=1.7, but you have torch 2.2.0 which is incompatible.\u001b[0m\n",
      "\u001b[34mSuccessfully installed Brotli-1.0.9 accelerate-0.33.0 bitsandbytes-0.39.1 black-23.7.0 datasets-2.14.1 docstring-parser-0.16 fire-0.5.0 huggingface-hub-0.24.2 inflate64-0.3.1 loralib-0.1.1 multivolumefile-0.2.3 mypy-extensions-1.0.0 nvidia-cublas-cu12-12.1.3.1 nvidia-cuda-cupti-cu12-12.1.105 nvidia-cuda-nvrtc-cu12-12.1.105 nvidia-cuda-runtime-cu12-12.1.105 nvidia-cudnn-cu12-8.9.2.26 nvidia-cufft-cu12-11.0.2.54 nvidia-curand-cu12-10.3.2.106 nvidia-cusolver-cu12-11.4.5.107 nvidia-cusparse-cu12-12.1.0.106 nvidia-nccl-cu12-2.19.3 nvidia-nvjitlink-cu12-12.3.101 nvidia-nvtx-cu12-12.1.105 pathspec-0.11.1 peft-0.4.0 py7zr-0.20.5 pybcj-1.0.1 pycryptodomex-3.18.0 pyppmd-1.0.0 pyzstd-0.15.9 safetensors-0.4.2 sagemaker-jumpstart-huggingface-script-utilities-1.2.5 sagemaker-jumpstart-script-utilities-1.1.9 shtab-1.7.1 termcolor-2.3.0 texttable-1.6.7 tokenize-rt-5.1.0 tokenizers-0.19.1 torch-2.2.0 transformers-4.43.1 triton-2.2.0 trl-0.8.1 typing-extensions-4.8.0 tyro-0.7.3\u001b[0m\n",
      "\u001b[34mWARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv\u001b[0m\n",
      "\u001b[34m2024-08-20 07:36:01,408 sagemaker-training-toolkit INFO     Waiting for the process to finish and give a return code.\u001b[0m\n",
      "\u001b[34m2024-08-20 07:36:01,408 sagemaker-training-toolkit INFO     Done waiting for a return code. Received 0 from exiting process.\u001b[0m\n",
      "\u001b[34m2024-08-20 07:36:01,449 sagemaker-training-toolkit INFO     No Neurons detected (normal if no neurons installed)\u001b[0m\n",
      "\u001b[34m2024-08-20 07:36:01,479 sagemaker-training-toolkit INFO     No Neurons detected (normal if no neurons installed)\u001b[0m\n",
      "\u001b[34m2024-08-20 07:36:01,507 sagemaker-training-toolkit INFO     No Neurons detected (normal if no neurons installed)\u001b[0m\n",
      "\u001b[34m2024-08-20 07:36:01,517 sagemaker-training-toolkit INFO     Invoking user script\u001b[0m\n",
      "\u001b[34mTraining Env:\u001b[0m\n",
      "\u001b[34m{\n",
      "    \"additional_framework_parameters\": {},\n",
      "    \"channel_input_dirs\": {\n",
      "        \"code\": \"/opt/ml/input/data/code\",\n",
      "        \"training\": \"/opt/ml/input/data/training\"\n",
      "    },\n",
      "    \"current_host\": \"algo-1\",\n",
      "    \"current_instance_group\": \"homogeneousCluster\",\n",
      "    \"current_instance_group_hosts\": [\n",
      "        \"algo-1\"\n",
      "    ],\n",
      "    \"current_instance_type\": \"ml.g5.2xlarge\",\n",
      "    \"distribution_hosts\": [],\n",
      "    \"distribution_instance_groups\": [],\n",
      "    \"framework_module\": \"sagemaker_pytorch_container.training:main\",\n",
      "    \"hosts\": [\n",
      "        \"algo-1\"\n",
      "    ],\n",
      "    \"hyperparameters\": {\n",
      "        \"add_input_output_demarcation_key\": \"True\",\n",
      "        \"chat_dataset\": \"False\",\n",
      "        \"enable_fsdp\": \"True\",\n",
      "        \"epoch\": \"5\",\n",
      "        \"instruction_tuned\": \"False\",\n",
      "        \"int8_quantization\": \"False\",\n",
      "        \"learning_rate\": \"0.0001\",\n",
      "        \"lora_alpha\": \"32\",\n",
      "        \"lora_dropout\": \"0.05\",\n",
      "        \"lora_r\": \"8\",\n",
      "        \"max_input_length\": \"-1\",\n",
      "        \"max_train_samples\": \"-1\",\n",
      "        \"max_val_samples\": \"-1\",\n",
      "        \"per_device_eval_batch_size\": \"1\",\n",
      "        \"per_device_train_batch_size\": \"4\",\n",
      "        \"preprocessing_num_workers\": \"None\",\n",
      "        \"seed\": \"10\",\n",
      "        \"target_modules\": \"q_proj,v_proj\",\n",
      "        \"train_data_split_seed\": \"0\",\n",
      "        \"validation_split_ratio\": \"0.2\"\n",
      "    },\n",
      "    \"input_config_dir\": \"/opt/ml/input/config\",\n",
      "    \"input_data_config\": {\n",
      "        \"code\": {\n",
      "            \"TrainingInputMode\": \"File\",\n",
      "            \"S3DistributionType\": \"FullyReplicated\",\n",
      "            \"RecordWrapperType\": \"None\"\n",
      "        },\n",
      "        \"training\": {\n",
      "            \"TrainingInputMode\": \"File\",\n",
      "            \"S3DistributionType\": \"FullyReplicated\",\n",
      "            \"RecordWrapperType\": \"None\"\n",
      "        }\n",
      "    },\n",
      "    \"input_dir\": \"/opt/ml/input\",\n",
      "    \"instance_groups\": [\n",
      "        \"homogeneousCluster\"\n",
      "    ],\n",
      "    \"instance_groups_dict\": {\n",
      "        \"homogeneousCluster\": {\n",
      "            \"instance_group_name\": \"homogeneousCluster\",\n",
      "            \"instance_type\": \"ml.g5.2xlarge\",\n",
      "            \"hosts\": [\n",
      "                \"algo-1\"\n",
      "            ]\n",
      "        }\n",
      "    },\n",
      "    \"is_hetero\": false,\n",
      "    \"is_master\": true,\n",
      "    \"is_modelparallel_enabled\": null,\n",
      "    \"is_smddpmprun_installed\": true,\n",
      "    \"job_name\": \"meta-textgeneration-llama-2-7b-2024-08-20-07-28-32-383\",\n",
      "    \"log_level\": 20,\n",
      "    \"master_hostname\": \"algo-1\",\n",
      "    \"model_dir\": \"/opt/ml/model\",\n",
      "    \"module_dir\": \"/opt/ml/input/data/code/sourcedir.tar.gz\",\n",
      "    \"module_name\": \"transfer_learning\",\n",
      "    \"network_interface_name\": \"eth0\",\n",
      "    \"num_cpus\": 8,\n",
      "    \"num_gpus\": 1,\n",
      "    \"num_neurons\": 0,\n",
      "    \"output_data_dir\": \"/opt/ml/output/data\",\n",
      "    \"output_dir\": \"/opt/ml/output\",\n",
      "    \"output_intermediate_dir\": \"/opt/ml/output/intermediate\",\n",
      "    \"resource_config\": {\n",
      "        \"current_host\": \"algo-1\",\n",
      "        \"current_instance_type\": \"ml.g5.2xlarge\",\n",
      "        \"current_group_name\": \"homogeneousCluster\",\n",
      "        \"hosts\": [\n",
      "            \"algo-1\"\n",
      "        ],\n",
      "        \"instance_groups\": [\n",
      "            {\n",
      "                \"instance_group_name\": \"homogeneousCluster\",\n",
      "                \"instance_type\": \"ml.g5.2xlarge\",\n",
      "                \"hosts\": [\n",
      "                    \"algo-1\"\n",
      "                ]\n",
      "            }\n",
      "        ],\n",
      "        \"network_interface_name\": \"eth0\"\n",
      "    },\n",
      "    \"user_entry_point\": \"transfer_learning.py\"\u001b[0m\n",
      "\u001b[34m}\u001b[0m\n",
      "\u001b[34mEnvironment variables:\u001b[0m\n",
      "\u001b[34mSM_HOSTS=[\"algo-1\"]\u001b[0m\n",
      "\u001b[34mSM_NETWORK_INTERFACE_NAME=eth0\u001b[0m\n",
      "\u001b[34mSM_HPS={\"add_input_output_demarcation_key\":\"True\",\"chat_dataset\":\"False\",\"enable_fsdp\":\"True\",\"epoch\":\"5\",\"instruction_tuned\":\"False\",\"int8_quantization\":\"False\",\"learning_rate\":\"0.0001\",\"lora_alpha\":\"32\",\"lora_dropout\":\"0.05\",\"lora_r\":\"8\",\"max_input_length\":\"-1\",\"max_train_samples\":\"-1\",\"max_val_samples\":\"-1\",\"per_device_eval_batch_size\":\"1\",\"per_device_train_batch_size\":\"4\",\"preprocessing_num_workers\":\"None\",\"seed\":\"10\",\"target_modules\":\"q_proj,v_proj\",\"train_data_split_seed\":\"0\",\"validation_split_ratio\":\"0.2\"}\u001b[0m\n",
      "\u001b[34mSM_USER_ENTRY_POINT=transfer_learning.py\u001b[0m\n",
      "\u001b[34mSM_FRAMEWORK_PARAMS={}\u001b[0m\n",
      "\u001b[34mSM_RESOURCE_CONFIG={\"current_group_name\":\"homogeneousCluster\",\"current_host\":\"algo-1\",\"current_instance_type\":\"ml.g5.2xlarge\",\"hosts\":[\"algo-1\"],\"instance_groups\":[{\"hosts\":[\"algo-1\"],\"instance_group_name\":\"homogeneousCluster\",\"instance_type\":\"ml.g5.2xlarge\"}],\"network_interface_name\":\"eth0\"}\u001b[0m\n",
      "\u001b[34mSM_INPUT_DATA_CONFIG={\"code\":{\"RecordWrapperType\":\"None\",\"S3DistributionType\":\"FullyReplicated\",\"TrainingInputMode\":\"File\"},\"training\":{\"RecordWrapperType\":\"None\",\"S3DistributionType\":\"FullyReplicated\",\"TrainingInputMode\":\"File\"}}\u001b[0m\n",
      "\u001b[34mSM_OUTPUT_DATA_DIR=/opt/ml/output/data\u001b[0m\n",
      "\u001b[34mSM_CHANNELS=[\"code\",\"training\"]\u001b[0m\n",
      "\u001b[34mSM_CURRENT_HOST=algo-1\u001b[0m\n",
      "\u001b[34mSM_CURRENT_INSTANCE_TYPE=ml.g5.2xlarge\u001b[0m\n",
      "\u001b[34mSM_CURRENT_INSTANCE_GROUP=homogeneousCluster\u001b[0m\n",
      "\u001b[34mSM_CURRENT_INSTANCE_GROUP_HOSTS=[\"algo-1\"]\u001b[0m\n",
      "\u001b[34mSM_INSTANCE_GROUPS=[\"homogeneousCluster\"]\u001b[0m\n",
      "\u001b[34mSM_INSTANCE_GROUPS_DICT={\"homogeneousCluster\":{\"hosts\":[\"algo-1\"],\"instance_group_name\":\"homogeneousCluster\",\"instance_type\":\"ml.g5.2xlarge\"}}\u001b[0m\n",
      "\u001b[34mSM_DISTRIBUTION_INSTANCE_GROUPS=[]\u001b[0m\n",
      "\u001b[34mSM_IS_HETERO=false\u001b[0m\n",
      "\u001b[34mSM_MODULE_NAME=transfer_learning\u001b[0m\n",
      "\u001b[34mSM_LOG_LEVEL=20\u001b[0m\n",
      "\u001b[34mSM_FRAMEWORK_MODULE=sagemaker_pytorch_container.training:main\u001b[0m\n",
      "\u001b[34mSM_INPUT_DIR=/opt/ml/input\u001b[0m\n",
      "\u001b[34mSM_INPUT_CONFIG_DIR=/opt/ml/input/config\u001b[0m\n",
      "\u001b[34mSM_OUTPUT_DIR=/opt/ml/output\u001b[0m\n",
      "\u001b[34mSM_NUM_CPUS=8\u001b[0m\n",
      "\u001b[34mSM_NUM_GPUS=1\u001b[0m\n",
      "\u001b[34mSM_NUM_NEURONS=0\u001b[0m\n",
      "\u001b[34mSM_MODEL_DIR=/opt/ml/model\u001b[0m\n",
      "\u001b[34mSM_MODULE_DIR=/opt/ml/input/data/code/sourcedir.tar.gz\u001b[0m\n",
      "\u001b[34mSM_TRAINING_ENV={\"additional_framework_parameters\":{},\"channel_input_dirs\":{\"code\":\"/opt/ml/input/data/code\",\"training\":\"/opt/ml/input/data/training\"},\"current_host\":\"algo-1\",\"current_instance_group\":\"homogeneousCluster\",\"current_instance_group_hosts\":[\"algo-1\"],\"current_instance_type\":\"ml.g5.2xlarge\",\"distribution_hosts\":[],\"distribution_instance_groups\":[],\"framework_module\":\"sagemaker_pytorch_container.training:main\",\"hosts\":[\"algo-1\"],\"hyperparameters\":{\"add_input_output_demarcation_key\":\"True\",\"chat_dataset\":\"False\",\"enable_fsdp\":\"True\",\"epoch\":\"5\",\"instruction_tuned\":\"False\",\"int8_quantization\":\"False\",\"learning_rate\":\"0.0001\",\"lora_alpha\":\"32\",\"lora_dropout\":\"0.05\",\"lora_r\":\"8\",\"max_input_length\":\"-1\",\"max_train_samples\":\"-1\",\"max_val_samples\":\"-1\",\"per_device_eval_batch_size\":\"1\",\"per_device_train_batch_size\":\"4\",\"preprocessing_num_workers\":\"None\",\"seed\":\"10\",\"target_modules\":\"q_proj,v_proj\",\"train_data_split_seed\":\"0\",\"validation_split_ratio\":\"0.2\"},\"input_config_dir\":\"/opt/ml/input/config\",\"input_data_config\":{\"code\":{\"RecordWrapperType\":\"None\",\"S3DistributionType\":\"FullyReplicated\",\"TrainingInputMode\":\"File\"},\"training\":{\"RecordWrapperType\":\"None\",\"S3DistributionType\":\"FullyReplicated\",\"TrainingInputMode\":\"File\"}},\"input_dir\":\"/opt/ml/input\",\"instance_groups\":[\"homogeneousCluster\"],\"instance_groups_dict\":{\"homogeneousCluster\":{\"hosts\":[\"algo-1\"],\"instance_group_name\":\"homogeneousCluster\",\"instance_type\":\"ml.g5.2xlarge\"}},\"is_hetero\":false,\"is_master\":true,\"is_modelparallel_enabled\":null,\"is_smddpmprun_installed\":true,\"job_name\":\"meta-textgeneration-llama-2-7b-2024-08-20-07-28-32-383\",\"log_level\":20,\"master_hostname\":\"algo-1\",\"model_dir\":\"/opt/ml/model\",\"module_dir\":\"/opt/ml/input/data/code/sourcedir.tar.gz\",\"module_name\":\"transfer_learning\",\"network_interface_name\":\"eth0\",\"num_cpus\":8,\"num_gpus\":1,\"num_neurons\":0,\"output_data_dir\":\"/opt/ml/output/data\",\"output_dir\":\"/opt/ml/output\",\"output_intermediate_dir\":\"/opt/ml/output/intermediate\",\"resource_config\":{\"current_group_name\":\"homogeneousCluster\",\"current_host\":\"algo-1\",\"current_instance_type\":\"ml.g5.2xlarge\",\"hosts\":[\"algo-1\"],\"instance_groups\":[{\"hosts\":[\"algo-1\"],\"instance_group_name\":\"homogeneousCluster\",\"instance_type\":\"ml.g5.2xlarge\"}],\"network_interface_name\":\"eth0\"},\"user_entry_point\":\"transfer_learning.py\"}\u001b[0m\n",
      "\u001b[34mSM_USER_ARGS=[\"--add_input_output_demarcation_key\",\"True\",\"--chat_dataset\",\"False\",\"--enable_fsdp\",\"True\",\"--epoch\",\"5\",\"--instruction_tuned\",\"False\",\"--int8_quantization\",\"False\",\"--learning_rate\",\"0.0001\",\"--lora_alpha\",\"32\",\"--lora_dropout\",\"0.05\",\"--lora_r\",\"8\",\"--max_input_length\",\"-1\",\"--max_train_samples\",\"-1\",\"--max_val_samples\",\"-1\",\"--per_device_eval_batch_size\",\"1\",\"--per_device_train_batch_size\",\"4\",\"--preprocessing_num_workers\",\"None\",\"--seed\",\"10\",\"--target_modules\",\"q_proj,v_proj\",\"--train_data_split_seed\",\"0\",\"--validation_split_ratio\",\"0.2\"]\u001b[0m\n",
      "\u001b[34mSM_OUTPUT_INTERMEDIATE_DIR=/opt/ml/output/intermediate\u001b[0m\n",
      "\u001b[34mSM_CHANNEL_CODE=/opt/ml/input/data/code\u001b[0m\n",
      "\u001b[34mSM_CHANNEL_TRAINING=/opt/ml/input/data/training\u001b[0m\n",
      "\u001b[34mSM_HP_ADD_INPUT_OUTPUT_DEMARCATION_KEY=True\u001b[0m\n",
      "\u001b[34mSM_HP_CHAT_DATASET=False\u001b[0m\n",
      "\u001b[34mSM_HP_ENABLE_FSDP=True\u001b[0m\n",
      "\u001b[34mSM_HP_EPOCH=5\u001b[0m\n",
      "\u001b[34mSM_HP_INSTRUCTION_TUNED=False\u001b[0m\n",
      "\u001b[34mSM_HP_INT8_QUANTIZATION=False\u001b[0m\n",
      "\u001b[34mSM_HP_LEARNING_RATE=0.0001\u001b[0m\n",
      "\u001b[34mSM_HP_LORA_ALPHA=32\u001b[0m\n",
      "\u001b[34mSM_HP_LORA_DROPOUT=0.05\u001b[0m\n",
      "\u001b[34mSM_HP_LORA_R=8\u001b[0m\n",
      "\u001b[34mSM_HP_MAX_INPUT_LENGTH=-1\u001b[0m\n",
      "\u001b[34mSM_HP_MAX_TRAIN_SAMPLES=-1\u001b[0m\n",
      "\u001b[34mSM_HP_MAX_VAL_SAMPLES=-1\u001b[0m\n",
      "\u001b[34mSM_HP_PER_DEVICE_EVAL_BATCH_SIZE=1\u001b[0m\n",
      "\u001b[34mSM_HP_PER_DEVICE_TRAIN_BATCH_SIZE=4\u001b[0m\n",
      "\u001b[34mSM_HP_PREPROCESSING_NUM_WORKERS=None\u001b[0m\n",
      "\u001b[34mSM_HP_SEED=10\u001b[0m\n",
      "\u001b[34mSM_HP_TARGET_MODULES=q_proj,v_proj\u001b[0m\n",
      "\u001b[34mSM_HP_TRAIN_DATA_SPLIT_SEED=0\u001b[0m\n",
      "\u001b[34mSM_HP_VALIDATION_SPLIT_RATIO=0.2\u001b[0m\n",
      "\u001b[34mPYTHONPATH=/opt/ml/code:/opt/conda/bin:/opt/conda/lib/python310.zip:/opt/conda/lib/python3.10:/opt/conda/lib/python3.10/lib-dynload:/opt/conda/lib/python3.10/site-packages\u001b[0m\n",
      "\u001b[34mInvoking script with the following command:\u001b[0m\n",
      "\u001b[34m/opt/conda/bin/python3.10 transfer_learning.py --add_input_output_demarcation_key True --chat_dataset False --enable_fsdp True --epoch 5 --instruction_tuned False --int8_quantization False --learning_rate 0.0001 --lora_alpha 32 --lora_dropout 0.05 --lora_r 8 --max_input_length -1 --max_train_samples -1 --max_val_samples -1 --per_device_eval_batch_size 1 --per_device_train_batch_size 4 --preprocessing_num_workers None --seed 10 --target_modules q_proj,v_proj --train_data_split_seed 0 --validation_split_ratio 0.2\u001b[0m\n",
      "\u001b[34m2024-08-20 07:36:01,561 sagemaker-training-toolkit INFO     Exceptions not imported for SageMaker TF as Tensorflow is not installed.\u001b[0m\n",
      "\u001b[34m===================================BUG REPORT===================================\u001b[0m\n",
      "\u001b[34mWelcome to bitsandbytes. For bug reports, please run\u001b[0m\n",
      "\u001b[34mpython -m bitsandbytes\n",
      " and submit this information together with your error trace to: https://github.com/TimDettmers/bitsandbytes/issues\u001b[0m\n",
      "\u001b[34m================================================================================\u001b[0m\n",
      "\u001b[34mbin /opt/conda/lib/python3.10/site-packages/bitsandbytes/libbitsandbytes_cuda118.so\u001b[0m\n",
      "\u001b[34m/opt/conda/lib/python3.10/site-packages/bitsandbytes/cuda_setup/main.py:149: UserWarning: WARNING: The following directories listed in your path were found to be non-existent: {PosixPath('/usr/local/nvidia/lib'), PosixPath('/usr/local/nvidia/lib64')}\n",
      "  warn(msg)\u001b[0m\n",
      "\u001b[34mCUDA SETUP: CUDA runtime path found: /opt/conda/lib/libcudart.so.11.0\u001b[0m\n",
      "\u001b[34mCUDA SETUP: Highest compute capability among GPUs detected: 8.6\u001b[0m\n",
      "\u001b[34mCUDA SETUP: Detected CUDA version 118\u001b[0m\n",
      "\u001b[34mCUDA SETUP: Loading binary /opt/conda/lib/python3.10/site-packages/bitsandbytes/libbitsandbytes_cuda118.so...\u001b[0m\n",
      "\u001b[34mINFO:root:Using pre-trained artifacts in SAGEMAKER_ADDITIONAL_S3_DATA_PATH=/opt/ml/additonals3data\u001b[0m\n",
      "\u001b[34mINFO:root:Identify file serving.properties in the un-tar directory /opt/ml/additonals3data. Copying it over to /opt/ml/model for model deployment after training is finished.\u001b[0m\n",
      "\u001b[34mINFO:root:Invoking the training command ['torchrun', '--nnodes', '1', '--nproc_per_node', '1', 'llama_finetuning.py', '--model_name', '/opt/ml/additonals3data', '--num_gpus', '1', '--pure_bf16', '--dist_checkpoint_root_folder', 'model_checkpoints', '--dist_checkpoint_folder', 'fine-tuned', '--batch_size_training', '4', '--micro_batch_size', '4', '--train_file', '/opt/ml/input/data/training', '--lr', '0.0001', '--do_train', '--output_dir', 'saved_peft_model', '--num_epochs', '5', '--use_peft', '--peft_method', 'lora', '--max_train_samples', '-1', '--max_val_samples', '-1', '--seed', '10', '--per_device_eval_batch_size', '1', '--max_input_length', '-1', '--preprocessing_num_workers', '--None', '--validation_split_ratio', '0.2', '--train_data_split_seed', '0', '--num_workers_dataloader', '0', '--weight_decay', '0.1', '--lora_r', '8', '--lora_alpha', '32', '--lora_dropout', '0.05', '--target_modules', 'q_proj,v_proj', '--enable_fsdp', '--add_input_output_demarcation_key'].\u001b[0m\n",
      "\u001b[34m===================================BUG REPORT===================================\u001b[0m\n",
      "\u001b[34mWelcome to bitsandbytes. For bug reports, please run\u001b[0m\n",
      "\u001b[34mpython -m bitsandbytes\n",
      " and submit this information together with your error trace to: https://github.com/TimDettmers/bitsandbytes/issues\u001b[0m\n",
      "\u001b[34m================================================================================\u001b[0m\n",
      "\u001b[34mbin /opt/conda/lib/python3.10/site-packages/bitsandbytes/libbitsandbytes_cuda118.so\u001b[0m\n",
      "\u001b[34m/opt/conda/lib/python3.10/site-packages/bitsandbytes/cuda_setup/main.py:149: UserWarning: WARNING: The following directories listed in your path were found to be non-existent: {PosixPath('/usr/local/nvidia/lib64'), PosixPath('/usr/local/nvidia/lib')}\n",
      "  warn(msg)\u001b[0m\n",
      "\u001b[34mCUDA SETUP: CUDA runtime path found: /opt/conda/lib/libcudart.so.11.0\u001b[0m\n",
      "\u001b[34mCUDA SETUP: Highest compute capability among GPUs detected: 8.6\u001b[0m\n",
      "\u001b[34mCUDA SETUP: Detected CUDA version 118\u001b[0m\n",
      "\u001b[34mCUDA SETUP: Loading binary /opt/conda/lib/python3.10/site-packages/bitsandbytes/libbitsandbytes_cuda118.so...\u001b[0m\n",
      "\u001b[34mINFO:root:Local rank is 0. Rank is 0\u001b[0m\n",
      "\u001b[34mINFO:root:Setting torch device = 0\u001b[0m\n",
      "\u001b[34mINFO:root:Loading the tokenizer.\u001b[0m\n",
      "\u001b[34m--> Running with torch dist debug set to detail\u001b[0m\n",
      "\u001b[34mINFO:root:'/opt/ml/additonals3data/__models_info__.json' file could not be found. Returning huggingface-model-id=info_not_present\u001b[0m\n",
      "\u001b[34mINFO:root:Loading the data.\u001b[0m\n",
      "\u001b[34mINFO:root:Both instruction_tuned and chat_dataset are set to False.Assuming domain adaptation dataset format.\u001b[0m\n",
      "\u001b[34mDownloading data files:   0%|          | 0/1 [00:00<?, ?it/s]\u001b[0m\n",
      "\u001b[34mDownloading data files: 100%|██████████| 1/1 [00:00<00:00, 14027.77it/s]\u001b[0m\n",
      "\u001b[34mExtracting data files:   0%|          | 0/1 [00:00<?, ?it/s]\u001b[0m\n",
      "\u001b[34mExtracting data files: 100%|██████████| 1/1 [00:00<00:00, 1309.49it/s]\u001b[0m\n",
      "\u001b[34mGenerating train split: 0 examples [00:00, ? examples/s]\u001b[0m\n",
      "\u001b[34mGenerating train split: 113 examples [00:00, 42760.41 examples/s]\u001b[0m\n",
      "\u001b[34mINFO:jumpstart:Training data is identified. The corresponded column names are ['text'].\u001b[0m\n",
      "\u001b[34mWARNING:jumpstart:The tokenizer picked has a `model_max_length` (1000000000000000019884624838656) larger than maximum input length cap 1024. Picking 1024 instead.\u001b[0m\n",
      "\u001b[34mINFO:jumpstart:The max sequence length is set as 1024.\u001b[0m\n",
      "\u001b[34mRunning tokenizer on dataset:   0%|          | 0/113 [00:00<?, ? examples/s]\u001b[0m\n",
      "\u001b[34mRunning tokenizer on dataset: 100%|██████████| 113/113 [00:00<00:00, 9079.28 examples/s]\u001b[0m\n",
      "\u001b[34mGrouping texts in chunks of 1024:   0%|          | 0/113 [00:00<?, ? examples/s]\u001b[0m\n",
      "\u001b[34mGrouping texts in chunks of 1024: 100%|██████████| 113/113 [00:00<00:00, 5518.89 examples/s]\u001b[0m\n",
      "\u001b[34mINFO:jumpstart:Test data is not identified. Split the data into train and test data respectively.\u001b[0m\n",
      "\u001b[34mINFO:root:Loading the pre-trained model.\u001b[0m\n",
      "\u001b[34mLoading checkpoint shards:   0%|          | 0/2 [00:00<?, ?it/s]\u001b[0m\n",
      "\u001b[34mLoading checkpoint shards:  50%|█████     | 1/2 [00:27<00:27, 27.81s/it]\u001b[0m\n",
      "\u001b[34mLoading checkpoint shards: 100%|██████████| 2/2 [00:38<00:00, 17.64s/it]\u001b[0m\n",
      "\u001b[34mLoading checkpoint shards: 100%|██████████| 2/2 [00:38<00:00, 19.17s/it]\u001b[0m\n",
      "\u001b[34m--> Model /opt/ml/additonals3data\u001b[0m\n",
      "\u001b[34m--> /opt/ml/additonals3data has 6738.415616 Million params\u001b[0m\n",
      "\u001b[34mtrainable params: 4,194,304 || all params: 6,742,609,920 || trainable%: 0.06220594176090199\u001b[0m\n",
      "\u001b[34mbFloat16 enabled for mixed precision - using bfSixteen policy\u001b[0m\n",
      "\u001b[34m--> applying fsdp activation checkpointing...\u001b[0m\n",
      "\u001b[34mINFO:root:--> Training Set Length = 8\u001b[0m\n",
      "\u001b[34mINFO:root:--> Validation Set Length = 2\u001b[0m\n",
      "\u001b[34m/opt/conda/lib/python3.10/site-packages/torch/cuda/memory.py:330: FutureWarning: torch.cuda.reset_max_memory_allocated now calls torch.cuda.reset_peak_memory_stats, which resets /all/ peak memory stats.\n",
      "  warnings.warn(\u001b[0m\n",
      "\u001b[34mTraining Epoch0:   0%|#033[34m          #033[0m| 0/2 [00:00<?, ?it/s]\u001b[0m\n",
      "\u001b[34m`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.\u001b[0m\n",
      "\u001b[34mNCCL version 2.19.3+cuda12.3\u001b[0m\n",
      "\u001b[34malgo-1:57:79 [0] nccl_net_ofi_init:1444 NCCL WARN NET/OFI Only EFA provider is supported\u001b[0m\n",
      "\u001b[34malgo-1:57:79 [0] nccl_net_ofi_init:1483 NCCL WARN NET/OFI aws-ofi-nccl initialization failed\u001b[0m\n",
      "\u001b[34mstep 0 is completed and loss is 2.1506943702697754\u001b[0m\n",
      "\u001b[34mTraining Epoch0:  50%|#033[34m█████     #033[0m| 1/2 [00:04<00:04,  4.93s/it]\u001b[0m\n",
      "\u001b[34mstep 1 is completed and loss is 2.0608088970184326\u001b[0m\n",
      "\u001b[34mTraining Epoch0: 100%|#033[34m██████████#033[0m| 2/2 [00:08<00:00,  4.06s/it]\u001b[0m\n",
      "\u001b[34mTraining Epoch0: 100%|#033[34m██████████#033[0m| 2/2 [00:08<00:00,  4.19s/it]\u001b[0m\n",
      "\u001b[34mMax CUDA memory allocated was 15 GB\u001b[0m\n",
      "\u001b[34mMax CUDA memory reserved was 15 GB\u001b[0m\n",
      "\u001b[34mPeak active CUDA memory was 15 GB\u001b[0m\n",
      "\u001b[34mCuda Malloc retires : 0\u001b[0m\n",
      "\u001b[34mCPU Total Peak Memory consumed during the train (max): 1 GB\u001b[0m\n",
      "\u001b[34mevaluating Epoch:   0%|#033[32m          #033[0m| 0/2 [00:00<?, ?it/s]\u001b[0m\n",
      "\u001b[34mWe detected that you are passing `past_key_values` as a tuple and this is deprecated and will be removed in v4.43. Please use an appropriate `Cache` class (https://huggingface.co/docs/transformers/v4.41.3/en/internal/generation_utils#transformers.Cache)\u001b[0m\n",
      "\u001b[34mevaluating Epoch:  50%|#033[32m█████     #033[0m| 1/2 [00:00<00:00,  3.07it/s]\u001b[0m\n",
      "\u001b[34mevaluating Epoch: 100%|#033[32m██████████#033[0m| 2/2 [00:00<00:00,  3.08it/s]\u001b[0m\n",
      "\u001b[34mevaluating Epoch: 100%|#033[32m██████████#033[0m| 2/2 [00:00<00:00,  3.08it/s]\u001b[0m\n",
      "\u001b[34meval_ppl=tensor(12.3573, device='cuda:0') eval_epoch_loss=tensor(2.5142, device='cuda:0')\u001b[0m\n",
      "\u001b[34mwe are about to save the PEFT modules\u001b[0m\n",
      "\u001b[34mPEFT modules are saved in saved_peft_model directory\u001b[0m\n",
      "\u001b[34mbest eval loss on epoch 0 is 2.5142431259155273\u001b[0m\n",
      "\u001b[34mEpoch 1: train_perplexity=8.2133, train_epoch_loss=2.1058, epcoh time 8.688848941999993s\u001b[0m\n",
      "\u001b[34mTraining Epoch1:   0%|#033[34m          #033[0m| 0/2 [00:00<?, ?it/s]\u001b[0m\n",
      "\u001b[34mstep 0 is completed and loss is 2.134112596511841\u001b[0m\n",
      "\u001b[34mTraining Epoch1:  50%|#033[34m█████     #033[0m| 1/2 [00:03<00:03,  3.45s/it]\u001b[0m\n",
      "\u001b[34mstep 1 is completed and loss is 2.0461034774780273\u001b[0m\n",
      "\u001b[34mTraining Epoch1: 100%|#033[34m██████████#033[0m| 2/2 [00:06<00:00,  3.45s/it]\u001b[0m\n",
      "\u001b[34mTraining Epoch1: 100%|#033[34m██████████#033[0m| 2/2 [00:06<00:00,  3.45s/it]\u001b[0m\n",
      "\u001b[34mMax CUDA memory allocated was 15 GB\u001b[0m\n",
      "\u001b[34mMax CUDA memory reserved was 15 GB\u001b[0m\n",
      "\u001b[34mPeak active CUDA memory was 15 GB\u001b[0m\n",
      "\u001b[34mCuda Malloc retires : 64\u001b[0m\n",
      "\u001b[34mCPU Total Peak Memory consumed during the train (max): 2 GB\u001b[0m\n",
      "\u001b[34mevaluating Epoch:   0%|#033[32m          #033[0m| 0/2 [00:00<?, ?it/s]\u001b[0m\n",
      "\u001b[34mevaluating Epoch:  50%|#033[32m█████     #033[0m| 1/2 [00:00<00:00,  3.11it/s]\u001b[0m\n",
      "\u001b[34mevaluating Epoch: 100%|#033[32m██████████#033[0m| 2/2 [00:00<00:00,  3.12it/s]\u001b[0m\n",
      "\u001b[34mevaluating Epoch: 100%|#033[32m██████████#033[0m| 2/2 [00:00<00:00,  3.11it/s]\u001b[0m\n",
      "\u001b[34meval_ppl=tensor(12.1266, device='cuda:0') eval_epoch_loss=tensor(2.4954, device='cuda:0')\u001b[0m\n",
      "\u001b[34mwe are about to save the PEFT modules\u001b[0m\n",
      "\u001b[34mPEFT modules are saved in saved_peft_model directory\u001b[0m\n",
      "\u001b[34mbest eval loss on epoch 1 is 2.495403289794922\u001b[0m\n",
      "\u001b[34mEpoch 2: train_perplexity=8.0858, train_epoch_loss=2.0901, epcoh time 7.367445795000037s\u001b[0m\n",
      "\u001b[34mTraining Epoch2:   0%|#033[34m          #033[0m| 0/2 [00:00<?, ?it/s]\u001b[0m\n",
      "\u001b[34mstep 0 is completed and loss is 2.1160759925842285\u001b[0m\n",
      "\u001b[34mTraining Epoch2:  50%|#033[34m█████     #033[0m| 1/2 [00:03<00:03,  3.45s/it]\u001b[0m\n",
      "\u001b[34mstep 1 is completed and loss is 2.027783155441284\u001b[0m\n",
      "\u001b[34mTraining Epoch2: 100%|#033[34m██████████#033[0m| 2/2 [00:06<00:00,  3.45s/it]\u001b[0m\n",
      "\u001b[34mTraining Epoch2: 100%|#033[34m██████████#033[0m| 2/2 [00:06<00:00,  3.45s/it]\u001b[0m\n",
      "\u001b[34mMax CUDA memory allocated was 15 GB\u001b[0m\n",
      "\u001b[34mMax CUDA memory reserved was 15 GB\u001b[0m\n",
      "\u001b[34mPeak active CUDA memory was 15 GB\u001b[0m\n",
      "\u001b[34mCuda Malloc retires : 128\u001b[0m\n",
      "\u001b[34mCPU Total Peak Memory consumed during the train (max): 2 GB\u001b[0m\n",
      "\u001b[34mevaluating Epoch:   0%|#033[32m          #033[0m| 0/2 [00:00<?, ?it/s]\u001b[0m\n",
      "\u001b[34mevaluating Epoch:  50%|#033[32m█████     #033[0m| 1/2 [00:00<00:00,  3.10it/s]\u001b[0m\n",
      "\u001b[34mevaluating Epoch: 100%|#033[32m██████████#033[0m| 2/2 [00:00<00:00,  3.12it/s]\u001b[0m\n",
      "\u001b[34mevaluating Epoch: 100%|#033[32m██████████#033[0m| 2/2 [00:00<00:00,  3.11it/s]\u001b[0m\n",
      "\u001b[34meval_ppl=tensor(11.8968, device='cuda:0') eval_epoch_loss=tensor(2.4763, device='cuda:0')\u001b[0m\n",
      "\u001b[34mwe are about to save the PEFT modules\u001b[0m\n",
      "\u001b[34mPEFT modules are saved in saved_peft_model directory\u001b[0m\n",
      "\u001b[34mbest eval loss on epoch 2 is 2.476273536682129\u001b[0m\n",
      "\u001b[34mEpoch 3: train_perplexity=7.9401, train_epoch_loss=2.0719, epcoh time 7.364584755999999s\u001b[0m\n",
      "\u001b[34mTraining Epoch3:   0%|#033[34m          #033[0m| 0/2 [00:00<?, ?it/s]\u001b[0m\n",
      "\u001b[34mstep 0 is completed and loss is 2.096951723098755\u001b[0m\n",
      "\u001b[34mTraining Epoch3:  50%|#033[34m█████     #033[0m| 1/2 [00:03<00:03,  3.45s/it]\u001b[0m\n",
      "\u001b[34mstep 1 is completed and loss is 2.007671356201172\u001b[0m\n",
      "\u001b[34mTraining Epoch3: 100%|#033[34m██████████#033[0m| 2/2 [00:06<00:00,  3.45s/it]\u001b[0m\n",
      "\u001b[34mTraining Epoch3: 100%|#033[34m██████████#033[0m| 2/2 [00:06<00:00,  3.45s/it]\u001b[0m\n",
      "\u001b[34mMax CUDA memory allocated was 15 GB\u001b[0m\n",
      "\u001b[34mMax CUDA memory reserved was 15 GB\u001b[0m\n",
      "\u001b[34mPeak active CUDA memory was 15 GB\u001b[0m\n",
      "\u001b[34mCuda Malloc retires : 192\u001b[0m\n",
      "\u001b[34mCPU Total Peak Memory consumed during the train (max): 2 GB\u001b[0m\n",
      "\u001b[34mevaluating Epoch:   0%|#033[32m          #033[0m| 0/2 [00:00<?, ?it/s]\u001b[0m\n",
      "\u001b[34mevaluating Epoch:  50%|#033[32m█████     #033[0m| 1/2 [00:00<00:00,  3.10it/s]\u001b[0m\n",
      "\u001b[34mevaluating Epoch: 100%|#033[32m██████████#033[0m| 2/2 [00:00<00:00,  3.12it/s]\u001b[0m\n",
      "\u001b[34mevaluating Epoch: 100%|#033[32m██████████#033[0m| 2/2 [00:00<00:00,  3.11it/s]\u001b[0m\n",
      "\u001b[34meval_ppl=tensor(11.6860, device='cuda:0') eval_epoch_loss=tensor(2.4584, device='cuda:0')\u001b[0m\n",
      "\u001b[34mwe are about to save the PEFT modules\u001b[0m\n",
      "\u001b[34mPEFT modules are saved in saved_peft_model directory\u001b[0m\n",
      "\u001b[34mbest eval loss on epoch 3 is 2.4583892822265625\u001b[0m\n",
      "\u001b[34mEpoch 4: train_perplexity=7.7859, train_epoch_loss=2.0523, epcoh time 7.366030215999899s\u001b[0m\n",
      "\u001b[34mTraining Epoch4:   0%|#033[34m          #033[0m| 0/2 [00:00<?, ?it/s]\u001b[0m\n",
      "\u001b[34mstep 0 is completed and loss is 2.077463150024414\u001b[0m\n",
      "\u001b[34mTraining Epoch4:  50%|#033[34m█████     #033[0m| 1/2 [00:03<00:03,  3.45s/it]\u001b[0m\n",
      "\u001b[34mstep 1 is completed and loss is 1.9897156953811646\u001b[0m\n",
      "\u001b[34mTraining Epoch4: 100%|#033[34m██████████#033[0m| 2/2 [00:06<00:00,  3.45s/it]\u001b[0m\n",
      "\u001b[34mTraining Epoch4: 100%|#033[34m██████████#033[0m| 2/2 [00:06<00:00,  3.45s/it]\u001b[0m\n",
      "\u001b[34mMax CUDA memory allocated was 15 GB\u001b[0m\n",
      "\u001b[34mMax CUDA memory reserved was 15 GB\u001b[0m\n",
      "\u001b[34mPeak active CUDA memory was 15 GB\u001b[0m\n",
      "\u001b[34mCuda Malloc retires : 257\u001b[0m\n",
      "\u001b[34mCPU Total Peak Memory consumed during the train (max): 2 GB\u001b[0m\n",
      "\u001b[34mevaluating Epoch:   0%|#033[32m          #033[0m| 0/2 [00:00<?, ?it/s]\u001b[0m\n",
      "\u001b[34mevaluating Epoch:  50%|#033[32m█████     #033[0m| 1/2 [00:00<00:00,  3.10it/s]\u001b[0m\n",
      "\u001b[34mevaluating Epoch: 100%|#033[32m██████████#033[0m| 2/2 [00:00<00:00,  3.11it/s]\u001b[0m\n",
      "\u001b[34mevaluating Epoch: 100%|#033[32m██████████#033[0m| 2/2 [00:00<00:00,  3.11it/s]\u001b[0m\n",
      "\u001b[34meval_ppl=tensor(11.4984, device='cuda:0') eval_epoch_loss=tensor(2.4422, device='cuda:0')\u001b[0m\n",
      "\u001b[34mwe are about to save the PEFT modules\u001b[0m\n",
      "\u001b[34mPEFT modules are saved in saved_peft_model directory\u001b[0m\n",
      "\u001b[34mbest eval loss on epoch 4 is 2.442208766937256\u001b[0m\n",
      "\u001b[34mEpoch 5: train_perplexity=7.6415, train_epoch_loss=2.0336, epcoh time 7.365390206000029s\u001b[0m\n",
      "\u001b[34mINFO:root:Key: avg_train_prep, Value: 7.9333062171936035\u001b[0m\n",
      "\u001b[34mINFO:root:Key: avg_train_loss, Value: 2.0707380771636963\u001b[0m\n",
      "\u001b[34mINFO:root:Key: avg_eval_prep, Value: 11.9130220413208\u001b[0m\n",
      "\u001b[34mINFO:root:Key: avg_eval_loss, Value: 2.4773037433624268\u001b[0m\n",
      "\u001b[34mINFO:root:Key: avg_epoch_time, Value: 7.630459982999992\u001b[0m\n",
      "\u001b[34mINFO:root:Key: avg_checkpoint_time, Value: 0.8130753604000347\u001b[0m\n",
      "\u001b[34mINFO:root:Combining pre-trained base model with the PEFT adapter module.\u001b[0m\n",
      "\u001b[34mLoading checkpoint shards:   0%|          | 0/2 [00:00<?, ?it/s]\u001b[0m\n",
      "\u001b[34mLoading checkpoint shards:  50%|█████     | 1/2 [00:29<00:29, 29.76s/it]\u001b[0m\n",
      "\u001b[34mLoading checkpoint shards: 100%|██████████| 2/2 [00:35<00:00, 15.48s/it]\u001b[0m\n",
      "\u001b[34mLoading checkpoint shards: 100%|██████████| 2/2 [00:35<00:00, 17.62s/it]\u001b[0m\n",
      "\u001b[34mINFO:root:Saving the combined model in safetensors format.\u001b[0m\n",
      "\u001b[34mINFO:root:Saving complete.\u001b[0m\n",
      "\u001b[34mINFO:root:Copying tokenizer to the output directory.\u001b[0m\n",
      "\u001b[34mINFO:root:Putting inference code with the fine-tuned model directory.\u001b[0m\n",
      "\u001b[34m2024-08-20 07:40:25,223 sagemaker-training-toolkit INFO     Waiting for the process to finish and give a return code.\u001b[0m\n",
      "\u001b[34m2024-08-20 07:40:25,223 sagemaker-training-toolkit INFO     Done waiting for a return code. Received 0 from exiting process.\u001b[0m\n",
      "\u001b[34m2024-08-20 07:40:25,224 sagemaker-training-toolkit INFO     Reporting training SUCCESS\u001b[0m\n",
      "\n",
      "2024-08-20 07:40:34 Uploading - Uploading generated training model\n",
      "2024-08-20 07:41:17 Completed - Training job completed\n",
      "Training seconds: 717\n",
      "Billable seconds: 717\n"
     ]
    }
   ],
   "source": [
    "from sagemaker.jumpstart.estimator import JumpStartEstimator\n",
    "import boto3\n",
    "\n",
    "estimator = JumpStartEstimator(model_id=model_id,  environment={\"accept_eula\": \"true\"},instance_type = \"ml.g5.2xlarge\") \n",
    "\n",
    "estimator.set_hyperparameters(instruction_tuned=\"False\", epoch=\"5\")\n",
    "\n",
    "#Fill in the code below with the dataset you want to use from above \n",
    "#example: estimator.fit({\"training\": f\"s3://genaiwithawsproject2024/training-datasets/finance\"})\n",
    "estimator.fit({\"training\": f\"s3://genaiwithawsproject2024/training-datasets/medical\"})"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Deploy the fine-tuned model\n",
    "---\n",
    "Next, we deploy the domain fine-tuned model. We will compare the performance of the fine-tuned and pre-trained model.\n",
    "\n",
    "---"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "No instance type selected for inference hosting endpoint. Defaulting to ml.g5.12xlarge.\n",
      "INFO:sagemaker.jumpstart:No instance type selected for inference hosting endpoint. Defaulting to ml.g5.12xlarge.\n",
      "INFO:sagemaker:Creating model with name: meta-textgeneration-llama-2-7b-2024-08-20-07-43-34-113\n",
      "INFO:sagemaker:Creating endpoint-config with name meta-textgeneration-llama-2-7b-2024-08-20-07-43-34-105\n",
      "INFO:sagemaker:Creating endpoint with name meta-textgeneration-llama-2-7b-2024-08-20-07-43-34-105\n"
     ]
    },
    {
      "name": "stdout",
      "output_type": "stream",
      "text": [
      "-------!"
     ]
    }
   ],
   "source": [
    "finetuned_predictor = estimator.deploy()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Evaluate the pre-trained and fine-tuned model\n",
    "---\n",
    "Next, we use the same input from the model evaluation step to evaluate the performance of the fine-tuned model and compare it with the base pre-trained model. \n",
    "\n",
    "---"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Create a function to print the response from the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "def print_response(payload, response):\n",
    "    print(payload[\"inputs\"])\n",
    "    print(f\"> {response}\")\n",
    "    print(\"\\n==================================\\n\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we can run the same prompts on the fine-tuned model to evaluate it's domain knowledge.   \n",
    "\n",
    "**Replace \"inputs\"** in the next cell with the input to send the model based on the domain you've chosen. \n",
    "\n",
    "**For financial domain:**\n",
    "\n",
    "  \"inputs\": \"Replace with sentence below from text\"  \n",
    "- \"The  investment  tests  performed  indicate\"\n",
    "- \"the  relative  volume  for  the  long  out  of  the  money  options, indicates\"\n",
    "- \"The  results  for  the  short  in  the  money  options\"\n",
    "- \"The  results  are  encouraging  for  aggressive  investors\"\n",
    "\n",
    "**For medical domain:** \n",
    "\n",
    "  \"inputs\": \"Replace with sentence below from text\" \n",
    "- \"Myeloid neoplasms and acute leukemias derive from\"\n",
    "- \"Genomic characterization is essential for\"\n",
    "- \"Certain germline disorders may be associated with\"\n",
    "- \"In contrast to targeted approaches, genome-wide sequencing\"\n",
    "\n",
    "**For IT domain:** \n",
    "\n",
    "  \"inputs\": \"Replace with sentence below from text\" \n",
    "- \"Traditional approaches to data management such as\"\n",
    "- \"A second important aspect of ubiquitous computing environments is\"\n",
    "- \"because ubiquitous computing is intended to\" \n",
    "- \"outline the key aspects of ubiquitous computing from a data management perspective.\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "tags": []
   },
   "outputs": [
      {
        "name": "stdout",
        "output_type": "stream",
        "text": [
          "Genomic characterization is essential for\n",
          "> [{'generated_text': ' the management of myeloid neoplasms and acute leukemia. It provides critical information that informs diagnosis, risk assessment, therapeutic decisions, and monitoring of disease progression and treatment resistance. \nAdvances in genomic techniques, including sequencing technologies and targeted gene panels, have significantly enhanced the accuracy'}]\n",
          "\n",
          "==================================\n",
          "\n"
        ]
      }
   ],
   "source": [
    "payload = {\n",
    "    \"inputs\": \"Genomic characterization is essential for\",\n",
    "    \"parameters\": {\n",
    "        \"max_new_tokens\": 64,\n",
    "        \"top_p\": 0.9,\n",
    "        \"temperature\": 0.6,\n",
    "        \"return_full_text\": False,\n",
    "    },\n",
    "}\n",
    "try:\n",
    "    response = finetuned_predictor.predict(payload, custom_attributes=\"accept_eula=true\")\n",
    "    print_response(payload, response)\n",
    "except Exception as e:\n",
    "    print(e)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Do the outputs from the fine-tuned model provide domain-specific insightful and relevant content? You can continue experimenting with the inputs of the model to test it's domain knowledge. \n",
    "\n",
    "**Use the output from this notebook to fill out the \"model fine-tuning\" section of the project documentation report**\n",
    "\n",
    "**After you've filled out the report, run the cells below to delete the model deployment** \n",
    "\n",
    "`IF YOU FAIL TO RUN THE CELLS BELOW YOU WILL RUN OUT OF BUDGET TO COMPLETE THE PROJECT`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "finetuned_predictor.delete_model()\n",
    "finetuned_predictor.delete_endpoint()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "conda_tensorflow2_p310",
   "language": "python",
   "name": "conda_tensorflow2_p310"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.14"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}