notebooks/2_tp_training_retrieval_model.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import time\n",
    "\n",
    "notebook_start_time = time.time()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Set up environment"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "⛳️ Local environment\n",
      "Adding the following directory to the PYTHONPATH: /Users/pauliusztin/Documents/01_projects/hopsworks_recsys/hands-on-recommender-system\n"
     ]
    }
   ],
   "source": [
    "import sys\n",
    "from pathlib import Path\n",
    "\n",
    "\n",
    "def is_google_colab() -> bool:\n",
    "    if \"google.colab\" in str(get_ipython()):\n",
    "        return True\n",
    "    return False\n",
    "\n",
    "\n",
    "def clone_repository() -> None:\n",
    "    !git clone https://github.com/decodingml/hands-on-recommender-system.git\n",
    "    %cd hands-on-recommender-system/\n",
    "\n",
    "\n",
    "def install_dependencies() -> None:\n",
    "    !pip install --upgrade uv\n",
    "    !uv pip install --all-extras --system --requirement pyproject.toml\n",
    "\n",
    "\n",
    "if is_google_colab():\n",
    "    clone_repository()\n",
    "    install_dependencies()\n",
    "\n",
    "    root_dir = str(Path().absolute())\n",
    "    print(\"⛳️ Google Colab environment\")\n",
    "else:\n",
    "    root_dir = str(Path().absolute().parent)\n",
    "    print(\"⛳️ Local environment\")\n",
    "\n",
    "# Add the root directory to the `PYTHONPATH` to use the `recsys` Python module from the notebook.\n",
    "if root_dir not in sys.path:\n",
    "    print(f\"Adding the following directory to the PYTHONPATH: {root_dir}\")\n",
    "    sys.path.append(root_dir)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 🧬 Training pipeline: Training retrieval model\n",
    "\n",
    "In this notebook, you will train a retrieval model that will be able to quickly generate a small subset of candidate items from a large collection of items. Your model will be based on the *two-tower architecture*, which embeds queries and candidates (keys) into a shared low-dimensional vector space. Here, a query consists of features of a customer and a transaction (e.g. timestamp of the purchase), whereas a candidate consists of features of a particular item. All queries will have a user ID and all candidates will have an item ID, and the model will be trained such that the embedding of a user will be close to all the embeddings of items the user has previously bought.\n",
    "\n",
    "After training the model you will save and upload its components to the Hopsworks Model Registry.\n",
    "\n",
    "Let's go ahead and load the data."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 📝 Imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "\n",
    "import warnings\n",
    "\n",
    "warnings.filterwarnings(\"ignore\")\n",
    "\n",
    "from loguru import logger\n",
    "\n",
    "from recsys import hopsworks_integration, training\n",
    "from recsys.config import settings"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Constants"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'CUSTOMER_DATA_SIZE': <CustomerDatasetSize.SMALL: 'SMALL'>,\n",
      " 'CUSTOM_HOPSWORKS_INFERENCE_ENV': 'custom_env_name',\n",
      " 'FEATURES_EMBEDDING_MODEL_ID': 'all-MiniLM-L6-v2',\n",
      " 'HOPSWORKS_API_KEY': SecretStr('**********'),\n",
      " 'OPENAI_API_KEY': SecretStr('**********'),\n",
      " 'OPENAI_MODEL_ID': 'gpt-4o-mini',\n",
      " 'RANKING_DATASET_VALIDATON_SPLIT_SIZE': 0.1,\n",
      " 'RANKING_EARLY_STOPPING_ROUNDS': 5,\n",
      " 'RANKING_ITERATIONS': 100,\n",
      " 'RANKING_LEARNING_RATE': 0.2,\n",
      " 'RANKING_MODEL_TYPE': 'ranking',\n",
      " 'RANKING_SCALE_POS_WEIGHT': 10,\n",
      " 'RECSYS_DIR': PosixPath('/Users/pauliusztin/Documents/01_projects/hopsworks_recsys/hands-on-recommender-system/recsys'),\n",
      " 'TWO_TOWER_DATASET_TEST_SPLIT_SIZE': 0.1,\n",
      " 'TWO_TOWER_DATASET_VALIDATON_SPLIT_SIZE': 0.1,\n",
      " 'TWO_TOWER_LEARNING_RATE': 0.01,\n",
      " 'TWO_TOWER_MODEL_BATCH_SIZE': 2048,\n",
      " 'TWO_TOWER_MODEL_EMBEDDING_SIZE': 16,\n",
      " 'TWO_TOWER_NUM_EPOCHS': 10,\n",
      " 'TWO_TOWER_WEIGHT_DECAY': 0.001}\n"
     ]
    }
   ],
   "source": [
    "from pprint import pprint\n",
    "\n",
    "pprint(dict(settings))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## <span style=\"color:#ff5f27\">🔮 Connect to Hopsworks Feature Store </span>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\u001b[32m2024-12-24 13:02:37.946\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mrecsys.hopsworks_integration.feature_store\u001b[0m:\u001b[36mget_feature_store\u001b[0m:\u001b[36m13\u001b[0m - \u001b[1mLoging to Hopsworks using HOPSWORKS_API_KEY env var.\u001b[0m\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2024-12-24 13:02:37,947 INFO: Initializing external client\n",
      "2024-12-24 13:02:37,947 INFO: Base URL: https://c.app.hopsworks.ai:443\n",
      "2024-12-24 13:02:39,358 INFO: Python Engine initialized.\n",
      "\n",
      "Logged in to project, explore it here https://c.app.hopsworks.ai:443/p/1192098\n"
     ]
    }
   ],
   "source": [
    "project, fs = hopsworks_integration.get_feature_store()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## <span style=\"color:#ff5f27\">Create Hopsworks Feature Views </span>\n",
    "\n",
    "In Hopsworks, you write features to feature groups (where the features are stored) and you read features from feature views. A feature view is a logical view over features, stored in feature groups, and a feature view typically contains the features used by a specific model. This way, feature views enable features, stored in different feature groups, to be reused across many different models.\n",
    "\n",
    "More in their [Feature Views](https://docs.hopsworks.ai/3.1/concepts/fs/feature_view/fv_overview/) docs."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "feature_view = hopsworks_integration.feature_store.create_retrieval_feature_view(fs)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "To view and explore data in the feature view you can retrieve batch data using the `get_batch_data()` method:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# feature_view.get_batch_data()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Create training dataset\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "You will train your retrieval model with a subset of features.\n",
    "\n",
    "For the query embedding you will use:\n",
    "- `customer_id`: ID of the customer.\n",
    "- `age`: age of the customer at the time of purchase.\n",
    "- `month_sin`, `month_cos`: time of year the purchase was made.\n",
    "\n",
    "For the candidate embedding you will use:\n",
    "- `article_id`: ID of the item.\n",
    "- `garment_group_name`: type of garment.\n",
    "- `index_group_name`: menswear/ladieswear etc."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\u001b[32m2024-12-24 13:02:45.199\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mrecsys.training.two_tower\u001b[0m:\u001b[36mget_train_val_split\u001b[0m:\u001b[36m263\u001b[0m - \u001b[1mRetrieving and creating train, val test split...\u001b[0m\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Finished: Reading data from Hopsworks, using Hopsworks Feature Query Service (5.22s) \n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "VersionWarning: Incremented version to `3`.\n"
     ]
    }
   ],
   "source": [
    "dataset = training.two_tower.TwoTowerDataset(\n",
    "    feature_view=feature_view, batch_size=settings.TWO_TOWER_MODEL_BATCH_SIZE\n",
    ")\n",
    "train_ds, val_ds = dataset.get_train_val_split()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's take a look at our dataset:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\u001b[32m2024-12-24 13:02:55.798\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36m__main__\u001b[0m:\u001b[36m<module>\u001b[0m:\u001b[36m1\u001b[0m - \u001b[1mTraining samples: 16,300\u001b[0m\n",
      "\u001b[32m2024-12-24 13:02:55.798\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36m__main__\u001b[0m:\u001b[36m<module>\u001b[0m:\u001b[36m2\u001b[0m - \u001b[1mValidation samples: 2,037\u001b[0m\n",
      "\u001b[32m2024-12-24 13:02:55.799\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36m__main__\u001b[0m:\u001b[36m<module>\u001b[0m:\u001b[36m4\u001b[0m - \u001b[1mNumber of users: 966\u001b[0m\n",
      "\u001b[32m2024-12-24 13:02:55.799\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36m__main__\u001b[0m:\u001b[36m<module>\u001b[0m:\u001b[36m5\u001b[0m - \u001b[1mNumber of items: 11,820\u001b[0m\n"
     ]
    }
   ],
   "source": [
    "logger.info(f\"Training samples: {len(dataset.properties['train_df']):,}\")\n",
    "logger.info(f\"Validation samples: {len(dataset.properties['val_df']):,}\")\n",
    "\n",
    "logger.info(f\"Number of users: {len(dataset.properties['user_ids']):,}\")\n",
    "logger.info(f\"Number of items: {len(dataset.properties['item_ids']):,}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>customer_id</th>\n",
       "      <th>article_id</th>\n",
       "      <th>t_dat</th>\n",
       "      <th>price</th>\n",
       "      <th>month_sin</th>\n",
       "      <th>month_cos</th>\n",
       "      <th>age</th>\n",
       "      <th>club_member_status</th>\n",
       "      <th>age_group</th>\n",
       "      <th>garment_group_name</th>\n",
       "      <th>index_group_name</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>f7048acb8188d98bde3a5c495475a3c86faafe0eede1f2...</td>\n",
       "      <td>670265002</td>\n",
       "      <td>0</td>\n",
       "      <td>0.013542</td>\n",
       "      <td>-8.660254e-01</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>48.0</td>\n",
       "      <td>ACTIVE</td>\n",
       "      <td>46-55</td>\n",
       "      <td>Under-, Nightwear</td>\n",
       "      <td>Ladieswear</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>5d34f84e6cbe9ec4706872bb65376097af1e53f0c7dac5...</td>\n",
       "      <td>751471035</td>\n",
       "      <td>0</td>\n",
       "      <td>0.033881</td>\n",
       "      <td>1.224647e-16</td>\n",
       "      <td>-1.000000</td>\n",
       "      <td>30.0</td>\n",
       "      <td>ACTIVE</td>\n",
       "      <td>26-35</td>\n",
       "      <td>Trousers</td>\n",
       "      <td>Ladieswear</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>baf6dc7ea8575732794751bb80824fe84fd40e6af86193...</td>\n",
       "      <td>719308002</td>\n",
       "      <td>0</td>\n",
       "      <td>0.059305</td>\n",
       "      <td>5.000000e-01</td>\n",
       "      <td>-0.866025</td>\n",
       "      <td>48.0</td>\n",
       "      <td>ACTIVE</td>\n",
       "      <td>46-55</td>\n",
       "      <td>Dresses Ladies</td>\n",
       "      <td>Divided</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>56f270317dfd89a08cb095acfd70778fda7770cdaa685c...</td>\n",
       "      <td>759231002</td>\n",
       "      <td>0</td>\n",
       "      <td>0.059305</td>\n",
       "      <td>1.224647e-16</td>\n",
       "      <td>-1.000000</td>\n",
       "      <td>28.0</td>\n",
       "      <td>ACTIVE</td>\n",
       "      <td>26-35</td>\n",
       "      <td>Dresses Ladies</td>\n",
       "      <td>Ladieswear</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>19ed12b3909e6d4fdb7e858c894d102e02d88691e98cf3...</td>\n",
       "      <td>793911001</td>\n",
       "      <td>0</td>\n",
       "      <td>0.033881</td>\n",
       "      <td>1.224647e-16</td>\n",
       "      <td>-1.000000</td>\n",
       "      <td>29.0</td>\n",
       "      <td>ACTIVE</td>\n",
       "      <td>26-35</td>\n",
       "      <td>Blouses</td>\n",
       "      <td>Ladieswear</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                         customer_id article_id  t_dat  \\\n",
       "0  f7048acb8188d98bde3a5c495475a3c86faafe0eede1f2...  670265002      0   \n",
       "1  5d34f84e6cbe9ec4706872bb65376097af1e53f0c7dac5...  751471035      0   \n",
       "2  baf6dc7ea8575732794751bb80824fe84fd40e6af86193...  719308002      0   \n",
       "3  56f270317dfd89a08cb095acfd70778fda7770cdaa685c...  759231002      0   \n",
       "5  19ed12b3909e6d4fdb7e858c894d102e02d88691e98cf3...  793911001      0   \n",
       "\n",
       "      price     month_sin  month_cos   age club_member_status age_group  \\\n",
       "0  0.013542 -8.660254e-01   0.500000  48.0             ACTIVE     46-55   \n",
       "1  0.033881  1.224647e-16  -1.000000  30.0             ACTIVE     26-35   \n",
       "2  0.059305  5.000000e-01  -0.866025  48.0             ACTIVE     46-55   \n",
       "3  0.059305  1.224647e-16  -1.000000  28.0             ACTIVE     26-35   \n",
       "5  0.033881  1.224647e-16  -1.000000  29.0             ACTIVE     26-35   \n",
       "\n",
       "  garment_group_name index_group_name  \n",
       "0  Under-, Nightwear       Ladieswear  \n",
       "1           Trousers       Ladieswear  \n",
       "2     Dresses Ladies          Divided  \n",
       "3     Dresses Ladies       Ladieswear  \n",
       "5            Blouses       Ladieswear  "
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset.properties[\"train_df\"].head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Build the Two Tower model\n",
    "\n",
    "The two tower model consist of two models:\n",
    "- **Query model**: Generates a query representation of a given user and transaction features.\n",
    "- **Candidate model**: Generates an item representation given item features.\n",
    "\n",
    "**Both models produce embeddings that live in the same embedding space**. You let this space be low-dimensional to prevent overfitting on the training data. (Otherwise, the model might simply memorize previous purchases, which makes it recommend items customers already have bought)."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "You start with creating the query model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "query_model_factory = training.two_tower.QueryTowerFactory(dataset=dataset)\n",
    "query_model = query_model_factory.build()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The candidate model is very similar to the query model. A difference is that it has two categorical features as input, which you one-hot encode."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "item_model_factory = training.two_tower.ItemTowerFactory(dataset=dataset)\n",
    "item_model = item_model_factory.build()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "You will evaluate the two tower model using the *top-100 accuracy*. That is, for each transaction in the validation data you will generate the associated query embedding and retrieve the set of the 100 items that are closest to this query in the embedding space. The top-100 accuracy measures how often the item that was actually bought is part of this subset. To evaluate this, you create a dataset of all unique items in the training data."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Create a TwoTowerModel with the specified **query_model** and **item_model**:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_factory = training.two_tower.TwoTowerFactory(dataset=dataset)\n",
    "model = model_factory.build(query_model=query_model, item_model=item_model)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Training the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2024-12-24 13:02:56,200 WARNING: At this time, the v2.11+ optimizer `tf.keras.optimizers.AdamW` runs slowly on M1/M2 Macs, please use the legacy Keras optimizer instead, located at `tf.keras.optimizers.legacy.AdamW`.\n",
      "Epoch 1/10\n",
      "8/8 [==============================] - 1s 25ms/step - loss: 15520.5253 - regularization_loss: 0.0000e+00 - total_loss: 15520.5253 - val_factorized_top_k/top_1_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_5_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_10_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_50_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_100_categorical_accuracy: 0.0000e+00 - val_loss: 15463.8848 - val_regularization_loss: 0.0000e+00 - val_total_loss: 15463.8848\n",
      "Epoch 2/10\n",
      "8/8 [==============================] - 0s 12ms/step - loss: 15319.1200 - regularization_loss: 0.0000e+00 - total_loss: 15319.1200 - val_factorized_top_k/top_1_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_5_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_10_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_50_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_100_categorical_accuracy: 0.0000e+00 - val_loss: 15334.2324 - val_regularization_loss: 0.0000e+00 - val_total_loss: 15334.2324\n",
      "Epoch 3/10\n",
      "8/8 [==============================] - 0s 12ms/step - loss: 14693.7700 - regularization_loss: 0.0000e+00 - total_loss: 14693.7700 - val_factorized_top_k/top_1_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_5_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_10_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_50_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_100_categorical_accuracy: 0.0000e+00 - val_loss: 15449.2861 - val_regularization_loss: 0.0000e+00 - val_total_loss: 15449.2861\n",
      "Epoch 4/10\n",
      "8/8 [==============================] - 0s 12ms/step - loss: 13692.2807 - regularization_loss: 0.0000e+00 - total_loss: 13692.2807 - val_factorized_top_k/top_1_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_5_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_10_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_50_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_100_categorical_accuracy: 0.0000e+00 - val_loss: 16580.0547 - val_regularization_loss: 0.0000e+00 - val_total_loss: 16580.0547\n",
      "Epoch 5/10\n",
      "8/8 [==============================] - 0s 12ms/step - loss: 12912.1939 - regularization_loss: 0.0000e+00 - total_loss: 12912.1939 - val_factorized_top_k/top_1_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_5_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_10_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_50_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_100_categorical_accuracy: 0.0000e+00 - val_loss: 16862.9121 - val_regularization_loss: 0.0000e+00 - val_total_loss: 16862.9121\n",
      "Epoch 6/10\n",
      "8/8 [==============================] - 0s 12ms/step - loss: 12178.2487 - regularization_loss: 0.0000e+00 - total_loss: 12178.2487 - val_factorized_top_k/top_1_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_5_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_10_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_50_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_100_categorical_accuracy: 0.0000e+00 - val_loss: 17456.7520 - val_regularization_loss: 0.0000e+00 - val_total_loss: 17456.7520\n",
      "Epoch 7/10\n",
      "8/8 [==============================] - 0s 12ms/step - loss: 11457.1168 - regularization_loss: 0.0000e+00 - total_loss: 11457.1168 - val_factorized_top_k/top_1_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_5_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_10_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_50_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_100_categorical_accuracy: 0.0000e+00 - val_loss: 19646.4336 - val_regularization_loss: 0.0000e+00 - val_total_loss: 19646.4336\n",
      "Epoch 8/10\n",
      "8/8 [==============================] - 0s 12ms/step - loss: 10679.1429 - regularization_loss: 0.0000e+00 - total_loss: 10679.1429 - val_factorized_top_k/top_1_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_5_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_10_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_50_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_100_categorical_accuracy: 0.0000e+00 - val_loss: 21145.1465 - val_regularization_loss: 0.0000e+00 - val_total_loss: 21145.1465\n",
      "Epoch 9/10\n",
      "8/8 [==============================] - 0s 12ms/step - loss: 10440.9469 - regularization_loss: 0.0000e+00 - total_loss: 10440.9469 - val_factorized_top_k/top_1_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_5_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_10_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_50_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_100_categorical_accuracy: 0.0000e+00 - val_loss: 23190.3633 - val_regularization_loss: 0.0000e+00 - val_total_loss: 23190.3633\n",
      "Epoch 10/10\n",
      "8/8 [==============================] - 0s 11ms/step - loss: 9795.1250 - regularization_loss: 0.0000e+00 - total_loss: 9795.1250 - val_factorized_top_k/top_1_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_5_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_10_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_50_categorical_accuracy: 0.0000e+00 - val_factorized_top_k/top_100_categorical_accuracy: 0.0000e+00 - val_loss: 23607.8770 - val_regularization_loss: 0.0000e+00 - val_total_loss: 23607.8770\n"
     ]
    }
   ],
   "source": [
    "trainer = training.two_tower.TwoTowerTrainer(dataset=dataset, model=model)\n",
    "history = trainer.train(train_ds, val_ds)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's take a look at the training and validation loss:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1000x600 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# Create figure with two subplots\n",
    "fig, (ax1, ax2) = plt.subplots(2, 1, figsize=(10, 6))\n",
    "\n",
    "# Training loss subplot\n",
    "ax1.plot(history.history[\"loss\"], label=\"Training Loss\", color=\"blue\")\n",
    "ax1.set_title(\"Training Loss Over Time\")\n",
    "ax1.set_xlabel(\"Epoch\")\n",
    "ax1.set_ylabel(\"Loss\")\n",
    "ax1.legend()\n",
    "ax1.grid(True)\n",
    "\n",
    "# Validation loss subplot\n",
    "ax2.plot(history.history[\"val_loss\"], label=\"Validation Loss\", color=\"red\")\n",
    "ax2.set_title(\"Validation Loss Over Time\")\n",
    "ax2.set_xlabel(\"Epoch\")\n",
    "ax2.set_ylabel(\"Loss\")\n",
    "ax2.legend()\n",
    "ax2.grid(True)\n",
    "\n",
    "# Adjust layout to prevent overlap\n",
    "plt.tight_layout()\n",
    "# plt.show() # Uncomment to show the plot"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# <span style=\"color:#ff5f27\">🗄️ Upload models to Hopsworks model registry </span>\n",
    "\n",
    "One of the features in Hopsworks is the model registry. This is where you can store different versions of models and compare their performance. Models from the registry can then be served as API endpoints.\n",
    "\n",
    "Let's connect to the model registry using the [HSML library](https://docs.hopsworks.ai/machine-learning-api/latest) from Hopsworks."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "mr = project.get_model_registry()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now, we push the models to Hopsworks model registry:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: query_model/assets\n",
      "2024-12-24 13:03:30,451 INFO: Assets written to: query_model/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Uploading: 100.000%|██████████| 55/55 elapsed<00:01 remaining<00:001<00:03,  1.29it/s]\n",
      "Uploading: 100.000%|██████████| 180107/180107 elapsed<00:02 remaining<00:00  1.29it/s]\n",
      "Uploading: 100.000%|██████████| 69498/69498 elapsed<00:01 remaining<00:003,  1.29it/s]\n",
      "Uploading: 100.000%|██████████| 561/561 elapsed<00:02 remaining<00:0000:03,  1.29it/s]\n",
      "Uploading: 100.000%|██████████| 152/152 elapsed<00:02 remaining<00:00<00:25,  6.32s/it]\n",
      "Model export complete: 100%|██████████| 6/6 [00:19<00:00,  3.19s/it]                   \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model created, explore it at https://c.app.hopsworks.ai:443/p/1192098/models/query_model/2\n",
      "INFO:tensorflow:Assets written to: candidate_model/assets\n",
      "2024-12-24 13:03:51,302 INFO: Assets written to: candidate_model/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Uploading: 100.000%|██████████| 55/55 elapsed<00:02 remaining<00:001<00:04,  1.20it/s]\n",
      "Uploading: 100.000%|██████████| 325891/325891 elapsed<00:02 remaining<00:00  1.20it/s]\n",
      "Uploading: 100.000%|██████████| 765738/765738 elapsed<00:04 remaining<00:00  1.20it/s]\n",
      "Uploading: 100.000%|██████████| 423/423 elapsed<00:03 remaining<00:0000:04,  1.20it/s]\n",
      "Uploading: 100.000%|██████████| 100/100 elapsed<00:02 remaining<00:00<00:36,  9.04s/it]\n",
      "Model export complete: 100%|██████████| 6/6 [00:24<00:00,  4.05s/it]                   "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model created, explore it at https://c.app.hopsworks.ai:443/p/1192098/models/candidate_model/2\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "query_model = hopsworks_integration.two_tower_serving.HopsworksQueryModel(\n",
    "    model=model.query_model\n",
    ")\n",
    "query_model.register(\n",
    "    mr=mr,\n",
    "    feature_view=feature_view,\n",
    "    query_df=dataset.properties[\"query_df\"],\n",
    ")\n",
    "\n",
    "item_model = hopsworks_integration.two_tower_serving.HopsworksCandidateModel(\n",
    "    model=model.item_model\n",
    ")\n",
    "item_model.register(\n",
    "    mr=mr,\n",
    "    feature_view=feature_view,\n",
    "    item_df=dataset.properties[\"item_df\"],\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## <span style=\"color:#ff5f27\"> Inspecting the models in the Hopsworks model registry </span>\n",
    "\n",
    "View results in [Hopsworks Serverless](https://rebrand.ly/serverless-github): **Data Science → Model Registry**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\u001b[32m2024-12-24 13:04:16.256\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36m__main__\u001b[0m:\u001b[36m<module>\u001b[0m:\u001b[36m4\u001b[0m - \u001b[1m⌛️ Notebook Execution time: 103.13 seconds ~ 1.72 minutes\u001b[0m\n"
     ]
    }
   ],
   "source": [
    "notebook_end_time = time.time()\n",
    "notebook_execution_time = notebook_end_time - notebook_start_time\n",
    "\n",
    "logger.info(\n",
    "    f\"⌛️ Notebook Execution time: {notebook_execution_time:.2f} seconds ~ {notebook_execution_time / 60:.2f} minutes\"\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# <span style=\"color:#ff5f27\">→ Next Steps </span>\n",
    "\n",
    "At this point you have a recommender system that is able to generate a set of candidate items for a customer. However, many of these could be poor, as the candidate model was trained with only a few subset of the features. In the next notebook, you'll train a *ranking model* to do more fine-grained predictions."
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.11.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}