create-llm-model.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Create Model and HF Model \n",
    "\n",
    "There are different way to create a LLM Hugging Model in SageMaker \n",
    "\n",
    "- Option 1. sm model + model uri + model data \n",
    "- Option 2. sm jumpstart \n",
    "- Option 3. HuggingFaceModel (can load model after endpoint created)\n",
    "- Option 4. The new HF LLM DLC get_huggingface_llm_image_uri\n",
    "\n",
    "- [sm rag docs and notebook](https://sagemaker-examples.readthedocs.io/en/latest/introduction_to_amazon_algorithms/jumpstart-foundation-models/question_answering_retrieval_augmented_generation/question_answering_jumpstart_knn.html)\n",
    "- [pgvector and sagemaker](https://github.com/aws-samples/aurora-postgresql-pgvector/blob/main/apgpgvector-similiarity-search/genai-pgvector-similarity-search.ipynb)\n",
    "- [hugging face hub model](https://www.philschmid.de/sagemaker-falcon-180b)\n",
    "- [Hugging Face LLM Inference Container ](https://www.philschmid.de/sagemaker-huggingface-llm)\n",
    "- [SageMaker HF Toolkit](https://github.com/aws/sagemaker-huggingface-inference-toolkit)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "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"
     ]
    }
   ],
   "source": [
    "from sagemaker import Session\n",
    "from sagemaker.model import Model \n",
    "from sagemaker.huggingface.model import HuggingFaceModel\n",
    "from sagemaker.huggingface import get_huggingface_llm_image_uri\n",
    "from sagemaker import image_uris\n",
    "from sagemaker import model_uris\n",
    "import json "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Parameters "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "endpoint_name = \"embedding_demo\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "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"
     ]
    }
   ],
   "source": [
    "session = Session()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_version = \"*\"\n",
    "model_id = \"huggingface-text2text-flan-t5-xxl\"\n",
    "instance_type = \"ml.g5.12xlarge\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Option 1. Create Model with Image and Model Data from Jumpstart "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# get image uri from sm  \n",
    "image_uri = image_uris.retrieve(\n",
    "  framework=None,\n",
    "  region=None,\n",
    "  image_scope=\"inference\",\n",
    "  model_id=model_id,\n",
    "  model_version=model_version,\n",
    "  instance_type=instance_type\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "763104351884.dkr.ecr.us-east-1.amazonaws.com/huggingface-pytorch-inference:1.13.1-transformers4.26.0-gpu-py39-cu117-ubuntu20.04\n"
     ]
    }
   ],
   "source": [
    "print(image_uri)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# get model data from sm jumpstart \n",
    "model_data = model_uris.retrieve(\n",
    "  region=None,\n",
    "  model_id=model_id, \n",
    "  model_version=model_version,\n",
    "  model_scope=\"inference\"\n",
    ")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "s3://jumpstart-cache-prod-us-east-1/huggingface-infer/prepack/v1.1.2/infer-prepack-huggingface-text2text-flan-t5-xxl.tar.gz\n"
     ]
    }
   ],
   "source": [
    "print(model_data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "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"
     ]
    }
   ],
   "source": [
    "model = Model(\n",
    "  image_uri=image_uri,\n",
    "  model_data=model_data,\n",
    "  role=None,\n",
    "  name=endpoint_name,\n",
    "  env={\"SAGEMAKER_MODEL_SERVER_WORKERS\": \"1\", \"TS_DEFAULT_WORKERS_PER_MODEL\": \"1\"}\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<sagemaker.model.Model object at 0x7fd18864c8b0>\n"
     ]
    }
   ],
   "source": [
    "print(model)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Option 2. Create Model with Image and Model Data From HF Hub "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "hub = {\n",
    "  'HF_MODEL_ID': 'sentence-transformers/all-MiniLM-L6-v2',\n",
    "  'HF_TASK': 'feature-extraction'\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "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",
      "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"
     ]
    }
   ],
   "source": [
    "hf_model = HuggingFaceModel(\n",
    "  role=None,\n",
    "  # modeal data could be loaded after endpoint created  \n",
    "  # model_data=model_data, \n",
    "  transformers_version=\"4.26\",\n",
    "  pytorch_version=\"1.13\",\n",
    "  py_version=\"py39\",\n",
    "  env=hub\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "None\n"
     ]
    }
   ],
   "source": [
    "print(hf_model.model_data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "None\n"
     ]
    }
   ],
   "source": [
    "print(hf_model.image_uri)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Option 3. Retrieve the new Hugging Face LLM DLC"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# get image uri from aws ecr \n",
    "# same hugging face llm dlc for different models \n",
    "llm_image_uri = get_huggingface_llm_image_uri(\n",
    "  session=session,\n",
    "  version=\"0.9.3\",\n",
    "  backend=\"huggingface\"\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "763104351884.dkr.ecr.us-east-1.amazonaws.com/huggingface-pytorch-tgi-inference:2.0.1-tgi0.9.3-gpu-py39-cu118-ubuntu20.04\n"
     ]
    }
   ],
   "source": [
    "print(llm_image_uri)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "hf_model_id = \"OpenAssistant/pythia-12b-sft-v8-7k-steps\" # model id from huggingface.co/models\n",
    "use_quantization = False # wether to use quantization or not\n",
    "instance_type = \"ml.g5.12xlarge\" # instance type to use for deployment\n",
    "number_of_gpu = 4 # number of gpus to use for inference and tensor parallelism\n",
    "health_check_timeout = 300 # Increase the timeout for the health check to 5 minutes for downloading the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "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",
      "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"
     ]
    }
   ],
   "source": [
    "llm_model = HuggingFaceModel(\n",
    "    role=None,\n",
    "    image_uri=llm_image_uri,\n",
    "    env={\n",
    "    'HF_MODEL_ID': hf_model_id,\n",
    "    'HF_MODEL_QUANTIZE': json.dumps(use_quantization),\n",
    "    'SM_NUM_GPUS': json.dumps(number_of_gpu)\n",
    "    }\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<sagemaker.huggingface.model.HuggingFaceModel object at 0x7f8aa2e68a30>\n"
     ]
    }
   ],
   "source": [
    "print(llm_model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "None\n"
     ]
    }
   ],
   "source": [
    "# model data will be loaded after the endpoint created \n",
    "print(llm_model.model_data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": ".env",
   "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": "3.9.17"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}