{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"name":"python","version":"3.11.13","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"},"kaggle":{"accelerator":"gpu","dataSources":[{"sourceId":38760,"databundleVersionId":4493939,"sourceType":"competition"},{"sourceId":4991874,"sourceType":"datasetVersion","datasetId":2895394}],"dockerImageVersionId":31193,"isInternetEnabled":true,"language":"python","sourceType":"notebook","isGpuEnabled":true}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"code","source":"!pip uninstall -y tensorflow tensorflow-intel tensorboard tensorboard-data-server tensorboard-plugin-wit\n!pip install -q recbole ray\n!pip install -q tqdm\n!pip install -q kmeans-pytorch\n!pip install -q orjson","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","trusted":true,"execution":{"iopub.status.busy":"2025-11-19T08:24:01.624297Z","iopub.execute_input":"2025-11-19T08:24:01.624487Z","iopub.status.idle":"2025-11-19T08:26:01.295061Z","shell.execute_reply.started":"2025-11-19T08:24:01.624461Z","shell.execute_reply":"2025-11-19T08:26:01.293983Z"}},"outputs":[{"name":"stdout","text":"Found existing installation: tensorflow 2.18.0\nUninstalling tensorflow-2.18.0:\n Successfully uninstalled tensorflow-2.18.0\n\u001b[33mWARNING: Skipping tensorflow-intel as it is not installed.\u001b[0m\u001b[33m\n\u001b[0mFound existing installation: tensorboard 2.18.0\nUninstalling tensorboard-2.18.0:\n Successfully uninstalled tensorboard-2.18.0\nFound existing installation: tensorboard-data-server 0.7.2\nUninstalling tensorboard-data-server-0.7.2:\n Successfully uninstalled tensorboard-data-server-0.7.2\n\u001b[33mWARNING: Skipping tensorboard-plugin-wit as it is not installed.\u001b[0m\u001b[33m\n\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m2.1/2.1 MB\u001b[0m \u001b[31m37.4 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m00:01\u001b[0m\n\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m57.1/57.1 MB\u001b[0m \u001b[31m31.9 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m:00:01\u001b[0m00:01\u001b[0m\n\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m5.5/5.5 MB\u001b[0m \u001b[31m97.8 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m:00:01\u001b[0m\n\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m363.4/363.4 MB\u001b[0m \u001b[31m4.7 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m:00:01\u001b[0m00:01\u001b[0m\n\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m13.8/13.8 MB\u001b[0m \u001b[31m93.7 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m:00:01\u001b[0m00:01\u001b[0m\n\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m24.6/24.6 MB\u001b[0m \u001b[31m76.8 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m:00:01\u001b[0m00:01\u001b[0m\n\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m883.7/883.7 kB\u001b[0m \u001b[31m44.1 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m664.8/664.8 MB\u001b[0m \u001b[31m2.5 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m:00:01\u001b[0m00:01\u001b[0m\n\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m211.5/211.5 MB\u001b[0m \u001b[31m5.3 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m:00:01\u001b[0m00:01\u001b[0m\n\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m56.3/56.3 MB\u001b[0m \u001b[31m29.6 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m:00:01\u001b[0m00:01\u001b[0m\n\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m127.9/127.9 MB\u001b[0m \u001b[31m9.7 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m:00:01\u001b[0m00:01\u001b[0m\n\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m207.5/207.5 MB\u001b[0m \u001b[31m2.2 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m:00:01\u001b[0m00:01\u001b[0m\n\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m21.1/21.1 MB\u001b[0m \u001b[31m83.6 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m:00:01\u001b[0m00:01\u001b[0m\n\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m6.6/6.6 MB\u001b[0m \u001b[31m100.4 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m00:01\u001b[0m\n\u001b[?25h\u001b[31mERROR: 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.\nbayesian-optimization 3.1.0 requires colorama>=0.4.6, but you have colorama 0.4.4 which is incompatible.\nlibcugraph-cu12 25.6.0 requires libraft-cu12==25.6.*, but you have libraft-cu12 25.2.0 which is incompatible.\npylibcugraph-cu12 25.6.0 requires pylibraft-cu12==25.6.*, but you have pylibraft-cu12 25.2.0 which is incompatible.\npylibcugraph-cu12 25.6.0 requires rmm-cu12==25.6.*, but you have rmm-cu12 25.2.0 which is incompatible.\u001b[0m\u001b[31m\n\u001b[0m","output_type":"stream"}],"execution_count":1},{"cell_type":"code","source":"import recbole\nimport torch\nfrom pathlib import Path\n\nprint(\"recbole version:\", recbole.__version__)\nprint(\"Torch version:\", torch.__version__)\n\nINPUT_DIR = Path(\"/kaggle/input/otto-recommender-system\")\nprint(\"Input dir exists:\", INPUT_DIR.exists())\nprint(\"Files:\", list(INPUT_DIR.iterdir()))\n","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-11-19T08:26:01.296292Z","iopub.execute_input":"2025-11-19T08:26:01.296596Z","iopub.status.idle":"2025-11-19T08:26:04.054257Z","shell.execute_reply.started":"2025-11-19T08:26:01.296565Z","shell.execute_reply":"2025-11-19T08:26:04.053568Z"}},"outputs":[{"name":"stdout","text":"recbole version: 1.2.1\nTorch version: 2.6.0+cu124\nInput dir exists: True\nFiles: [PosixPath('/kaggle/input/otto-recommender-system/sample_submission.csv'), PosixPath('/kaggle/input/otto-recommender-system/test.jsonl'), PosixPath('/kaggle/input/otto-recommender-system/train.jsonl')]\n","output_type":"stream"}],"execution_count":2},{"cell_type":"code","source":"import numpy as np\n\nif not hasattr(np, \"float\"):\n np.float = float\nif not hasattr(np, \"int\"):\n np.int = int\nif not hasattr(np, \"bool\"):\n np.bool = bool\nif not hasattr(np, \"object\"):\n np.object = object\nif not hasattr(np, \"complex\"):\n np.complex = complex\nif not hasattr(np, \"complex_\"):\n np.complex_ = np.complex128","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-11-19T08:26:04.05577Z","iopub.execute_input":"2025-11-19T08:26:04.056073Z","iopub.status.idle":"2025-11-19T08:26:04.061283Z","shell.execute_reply.started":"2025-11-19T08:26:04.056055Z","shell.execute_reply":"2025-11-19T08:26:04.060494Z"}},"outputs":[{"name":"stderr","text":"/tmp/ipykernel_48/2133633671.py:7: FutureWarning: In the future `np.bool` will be defined as the corresponding NumPy scalar.\n if not hasattr(np, \"bool\"):\n/tmp/ipykernel_48/2133633671.py:9: FutureWarning: In the future `np.object` will be defined as the corresponding NumPy scalar.\n if not hasattr(np, \"object\"):\n","output_type":"stream"}],"execution_count":3},{"cell_type":"code","source":"import os\nimport json\nfrom pathlib import Path\n\nimport pandas as pd\nimport numpy as np\nfrom tqdm import tqdm\n\nimport torch\n\nfrom recbole.quick_start import run_recbole\nfrom recbole.config import Config\nfrom recbole.data import create_dataset, data_preparation\nfrom recbole.model.sequential_recommender import BERT4Rec\nfrom recbole.trainer import Trainer\nfrom recbole.utils import init_seed","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-11-19T08:26:04.062153Z","iopub.execute_input":"2025-11-19T08:26:04.062444Z","iopub.status.idle":"2025-11-19T08:26:11.426589Z","shell.execute_reply.started":"2025-11-19T08:26:04.062426Z","shell.execute_reply":"2025-11-19T08:26:11.425787Z"}},"outputs":[],"execution_count":4},{"cell_type":"code","source":"# Test\nTEST_MODE = True \nN_TRAIN_SESS = 20000 \nN_TEST_SESS = 2000 \nEPOCHS = 1 \nHIDDEN_SIZE = 128 \nNUM_LAYERS = 2\nMAX_ITEM_LIST_LENGTH = 50\nTRAIN_BATCH_SIZE = 512","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-11-19T08:26:11.427452Z","iopub.execute_input":"2025-11-19T08:26:11.428031Z","iopub.status.idle":"2025-11-19T08:26:11.432277Z","shell.execute_reply.started":"2025-11-19T08:26:11.428006Z","shell.execute_reply":"2025-11-19T08:26:11.431571Z"}},"outputs":[],"execution_count":5},{"cell_type":"code","source":"INPUT_DIR = Path(\"/kaggle/input/otto-recommender-system\")\nprint(\"Input dir exists:\", INPUT_DIR.exists())\nprint(\"Files in input:\", list(INPUT_DIR.iterdir()))","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-11-19T08:26:11.432891Z","iopub.execute_input":"2025-11-19T08:26:11.433192Z","iopub.status.idle":"2025-11-19T08:26:11.453155Z","shell.execute_reply.started":"2025-11-19T08:26:11.433165Z","shell.execute_reply":"2025-11-19T08:26:11.452533Z"}},"outputs":[{"name":"stdout","text":"Input dir exists: True\nFiles in input: [PosixPath('/kaggle/input/otto-recommender-system/sample_submission.csv'), PosixPath('/kaggle/input/otto-recommender-system/test.jsonl'), PosixPath('/kaggle/input/otto-recommender-system/train.jsonl')]\n","output_type":"stream"}],"execution_count":6},{"cell_type":"code","source":"import orjson\nfrom pathlib import Path\nfrom tqdm import tqdm\n\n# ===== CẤU HÌNH =====\nINPUT_DIR = Path(\"/kaggle/input/otto-recommender-system\")\nTRAIN_JSONL = INPUT_DIR / \"train.jsonl\"\n\nSMALL_TRAIN = True # test: True, train thật: False\nN_TRAIN_SESS = 20_000\nEVENT_FILTER = \"all\"\n\nDATASET_ROOT = Path(\"./dataset\")\nDATASET_NAME = \"otto_small\" if SMALL_TRAIN else \"otto\"\n\n# Tạo thư mục chắc chắn tồn tại\nOTTO_DIR = DATASET_ROOT / DATASET_NAME\nOTTO_DIR.mkdir(parents=True, exist_ok=True)\n\ninter_path = OTTO_DIR / f\"{DATASET_NAME}.inter\"\n\nprint(\"TRAIN_JSONL exists:\", TRAIN_JSONL.exists())\nprint(\"Saving to:\", inter_path)\n\nnum_sessions = 0\nnum_events_total = 0\nnum_events_kept = 0\n\nwith open(TRAIN_JSONL, \"rb\") as fin, open(inter_path, \"w\") as fout:\n # header RecBole\n fout.write(\"user_id:token\\titem_id:token\\ttimestamp:float\\n\")\n\n for line in tqdm(\n fin,\n desc=f\"Writing {DATASET_NAME}.inter (streaming)\"\n ):\n if SMALL_TRAIN and num_sessions >= N_TRAIN_SESS:\n break\n\n obj = orjson.loads(line)\n session_id = obj[\"session\"]\n events = obj[\"events\"]\n num_sessions += 1\n num_events_total += len(events)\n\n for ev in events:\n if EVENT_FILTER == \"clicks\" and ev[\"type\"] != \"clicks\":\n continue\n fout.write(f\"{session_id}\\t{ev['aid']}\\t{float(ev['ts'])}\\n\")\n num_events_kept += 1\n\nprint(\"✅ Done\")\nprint(\"Sessions:\", num_sessions)\nprint(\"Events read:\", num_events_total)\nprint(\"Events written:\", num_events_kept)\nprint(\"File size (MB):\", inter_path.stat().st_size / (1024 * 1024))","metadata":{"execution":{"iopub.status.busy":"2025-11-19T08:26:11.453872Z","iopub.execute_input":"2025-11-19T08:26:11.454476Z","iopub.status.idle":"2025-11-19T08:26:13.57079Z","shell.execute_reply.started":"2025-11-19T08:26:11.454452Z","shell.execute_reply":"2025-11-19T08:26:13.570015Z"},"trusted":true},"outputs":[{"name":"stdout","text":"TRAIN_JSONL exists: True\nSaving to: dataset/otto_small/otto_small.inter\n","output_type":"stream"},{"name":"stderr","text":"Writing otto_small.inter (streaming): 20000it [00:02, 9543.82it/s] ","output_type":"stream"},{"name":"stdout","text":"✅ Done\nSessions: 20000\nEvents read: 1127647\nEvents written: 1127647\nFile size (MB): 30.93029022216797\n","output_type":"stream"},{"name":"stderr","text":"\n","output_type":"stream"}],"execution_count":7},{"cell_type":"code","source":"SMALL_TRAIN = True # 👉 test: True, train thật: False\n\nif SMALL_TRAIN:\n EPOCHS = 2\n MAX_LEN = 50\n HIDDEN_SIZE = 128\n NUM_LAYERS = 2\n TRAIN_BATCH = 128 # 🔻 GIẢM mạnh batch size\nelse:\n EPOCHS = 10\n MAX_LEN = 50\n HIDDEN_SIZE = 256\n NUM_LAYERS = 2\n TRAIN_BATCH = 256 # nếu vẫn OOM thì hạ tiếp xuống 128\n\nparameter_dict = {\n # --- Data ---\n \"data_path\": str(DATASET_ROOT),\n \"dataset\": DATASET_NAME,\n \"field_separator\": \"\\t\",\n \"load_col\": {\"inter\": [\"user_id\", \"item_id\", \"timestamp\"]},\n \"USER_ID_FIELD\": \"user_id\",\n \"ITEM_ID_FIELD\": \"item_id\",\n \"TIME_FIELD\": \"timestamp\",\n\n # --- Sequential ---\n \"MAX_ITEM_LIST_LENGTH\": MAX_LEN,\n\n # --- Model BERT4Rec ---\n \"hidden_size\": HIDDEN_SIZE,\n \"num_layers\": NUM_LAYERS,\n \"num_heads\": 4,\n \"dropout_prob\": 0.2,\n\n # --- Training ---\n \"epochs\": EPOCHS,\n \"train_batch_size\": TRAIN_BATCH,\n \"eval_batch_size\": 4096,\n \"learning_rate\": 1e-3,\n \"train_neg_sample_args\": None,\n\n # --- Evaluation ---\n \"eval_args\": {\n \"split\": {\"RS\": [0.99, 0.01, 0.0]},\n \"order\": \"TO\",\n \"mode\": \"full\"\n },\n \"metrics\": [\"Recall\", \"MRR\"],\n \"topk\": [20],\n \"valid_metric\": \"Recall@20\",\n\n # --- Logging ---\n \"show_progress\": True, # tqdm ON\n \"state\": \"WARNING\", # giảm log tối đa\n\n # --- Misc ---\n \"device\": \"cuda\" if torch.cuda.is_available() else \"cpu\",\n \"tensorboard\": False,\n \"reproducibility\": True,\n}","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-11-19T08:26:13.57162Z","iopub.execute_input":"2025-11-19T08:26:13.572092Z","iopub.status.idle":"2025-11-19T08:26:13.627799Z","shell.execute_reply.started":"2025-11-19T08:26:13.572071Z","shell.execute_reply":"2025-11-19T08:26:13.626922Z"}},"outputs":[],"execution_count":8},{"cell_type":"code","source":"'''if SMALL_TRAIN:\n # Lấy sample theo session để giữ nguyên cấu trúc chuỗi\n sample_sessions = train_df[\"session\"].drop_duplicates().head(N_TRAIN_SESS)\n train_df_small = train_df[train_df[\"session\"].isin(sample_sessions)].copy()\n print(\n \"Dùng sample train nhỏ:\",\n train_df_small[\"session\"].nunique(), \"sessions,\",\n len(train_df_small), \"interactions\"\n )\nelse:\n # Dùng toàn bộ dữ liệu đã load vào train_df\n train_df_small = train_df\n\ntrain_df_small.head()'''","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-11-19T08:26:13.631148Z","iopub.execute_input":"2025-11-19T08:26:13.631617Z","iopub.status.idle":"2025-11-19T08:26:13.646899Z","shell.execute_reply.started":"2025-11-19T08:26:13.631586Z","shell.execute_reply":"2025-11-19T08:26:13.646325Z"}},"outputs":[{"execution_count":9,"output_type":"execute_result","data":{"text/plain":"'if SMALL_TRAIN:\\n # Lấy sample theo session để giữ nguyên cấu trúc chuỗi\\n sample_sessions = train_df[\"session\"].drop_duplicates().head(N_TRAIN_SESS)\\n train_df_small = train_df[train_df[\"session\"].isin(sample_sessions)].copy()\\n print(\\n \"Dùng sample train nhỏ:\",\\n train_df_small[\"session\"].nunique(), \"sessions,\",\\n len(train_df_small), \"interactions\"\\n )\\nelse:\\n # Dùng toàn bộ dữ liệu đã load vào train_df\\n train_df_small = train_df\\n\\ntrain_df_small.head()'"},"metadata":{}}],"execution_count":9},{"cell_type":"code","source":"'''# Thư mục dataset phụ thuộc vào SMALL_TRAIN\nOTTO_DIR = DATASET_ROOT / DATASET_NAME\nOTTO_DIR.mkdir(parents=True, exist_ok=True)\n\ninter_path = OTTO_DIR / f\"{DATASET_NAME}.inter\"\nprint(\"Saving to:\", inter_path)\n\nwith open(inter_path, \"w\") as f:\n # header RecBole: field_name:type\n f.write(\"user_id:token\\titem_id:token\\ttimestamp:float\\n\")\n for row in tqdm(\n train_df_small[[\"session\", \"aid\", \"ts\"]].itertuples(index=False),\n desc=f\"Writing {DATASET_NAME}.inter\"\n ):\n # session = user_id, aid = item_id\n f.write(f\"{row.session}\\t{row.aid}\\t{float(row.ts)}\\n\")\n\nprint(\"File size (MB):\", inter_path.stat().st_size / (1024 * 1024))'''","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-11-19T08:26:13.647587Z","iopub.execute_input":"2025-11-19T08:26:13.647797Z","iopub.status.idle":"2025-11-19T08:26:13.664571Z","shell.execute_reply.started":"2025-11-19T08:26:13.647774Z","shell.execute_reply":"2025-11-19T08:26:13.663845Z"}},"outputs":[{"execution_count":10,"output_type":"execute_result","data":{"text/plain":"'# Thư mục dataset phụ thuộc vào SMALL_TRAIN\\nOTTO_DIR = DATASET_ROOT / DATASET_NAME\\nOTTO_DIR.mkdir(parents=True, exist_ok=True)\\n\\ninter_path = OTTO_DIR / f\"{DATASET_NAME}.inter\"\\nprint(\"Saving to:\", inter_path)\\n\\nwith open(inter_path, \"w\") as f:\\n # header RecBole: field_name:type\\n f.write(\"user_id:token\\titem_id:token\\ttimestamp:float\\n\")\\n for row in tqdm(\\n train_df_small[[\"session\", \"aid\", \"ts\"]].itertuples(index=False),\\n desc=f\"Writing {DATASET_NAME}.inter\"\\n ):\\n # session = user_id, aid = item_id\\n f.write(f\"{row.session}\\t{row.aid}\\t{float(row.ts)}\\n\")\\n\\nprint(\"File size (MB):\", inter_path.stat().st_size / (1024 * 1024))'"},"metadata":{}}],"execution_count":10},{"cell_type":"code","source":"from recbole.quick_start import run_recbole\n\nresults = run_recbole(\n model=\"BERT4Rec\",\n dataset=DATASET_NAME,\n config_dict=parameter_dict\n)\nprint(results)","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-11-19T08:26:13.665301Z","iopub.execute_input":"2025-11-19T08:26:13.665484Z"}},"outputs":[{"name":"stderr","text":"command line args [-f /root/.local/share/jupyter/runtime/kernel-039efa5c-941b-468e-88fd-191c3debbef2.json] will not be used in RecBole\n/usr/local/lib/python3.11/dist-packages/recbole/data/dataset/dataset.py:648: FutureWarning: A value is trying to be set on a copy of a DataFrame or Series through chained assignment using an inplace method.\nThe behavior will change in pandas 3.0. This inplace method will never work because the intermediate object on which we are setting values always behaves as a copy.\n\nFor example, when doing 'df[col].method(value, inplace=True)', try using 'df.method({col: value}, inplace=True)' or df[col] = df[col].method(value) instead, to perform the operation inplace on the original object.\n\n\n feat[field].fillna(value=0, inplace=True)\n/usr/local/lib/python3.11/dist-packages/recbole/data/dataset/dataset.py:650: FutureWarning: A value is trying to be set on a copy of a DataFrame or Series through chained assignment using an inplace method.\nThe behavior will change in pandas 3.0. This inplace method will never work because the intermediate object on which we are setting values always behaves as a copy.\n\nFor example, when doing 'df[col].method(value, inplace=True)', try using 'df.method({col: value}, inplace=True)' or df[col] = df[col].method(value) instead, to perform the operation inplace on the original object.\n\n\n feat[field].fillna(value=feat[field].mean(), inplace=True)\nTrain 0: 0%| | 0/8493 [00:00 0, \"Không tìm thấy checkpoint model BERT4Rec\"\nbest_model_path = model_ckpts[-1]\nbest_model_path","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"config = Config(\n model=\"BERT4Rec\",\n dataset=\"otto\",\n config_dict=parameter_dict\n)\n\ninit_seed(config[\"seed\"], config[\"reproducibility\"])\n\ndataset = create_dataset(config)\ntrain_data, valid_data, test_data = data_preparation(config, dataset)\n\nmodel = BERT4Rec(config, train_data.dataset).to(config[\"device\"])\nmodel.load_state_dict(\n torch.load(best_model_path, map_location=config[\"device\"])\n)\nmodel.eval()\n\nprint(\"Model loaded on:\", config[\"device\"])","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"test_path = INPUT_DIR / \"test.jsonl\"\n\ntest_sessions = []\nwith open(test_path, \"r\") as f:\n for line in tqdm(f, desc=\"Reading test.jsonl\"):\n obj = json.loads(line)\n session_id = obj[\"session\"]\n events = obj[\"events\"]\n events = sorted(events, key=lambda x: x[\"ts\"])\n aids = [e[\"aid\"] for e in events]\n test_sessions.append((session_id, aids))\n\nprint(\"Tổng số test sessions:\", len(test_sessions))\n\nif TEST_MODE:\n test_sessions = test_sessions[:N_TEST_SESS]\n print(\"Dùng sample test nhỏ:\", len(test_sessions), \"sessions\")\n\ntest_sessions[0]","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# Map từ raw token -> index và ngược lại\nuid2token = dataset.field2token_id[\"user_id\"]\niid2token = dataset.field2token_id[\"item_id\"]\n\ntoken2uid = {v: k for k, v in uid2token.items()}\ntoken2iid = {v: k for k, v in iid2token.items()}\n\nprint(\"Num users:\", len(uid2token), \"Num items:\", len(iid2token))","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"MAX_LEN = config[\"MAX_ITEM_LIST_LENGTH\"]\n\ndef build_input_batch(session_aids_batch):\n \"\"\"\n session_aids_batch: list[list[aid]]\n Trả về dict input batch cho model.full_sort_predict()\n \"\"\"\n seqs = []\n lengths = []\n\n for aids in session_aids_batch:\n # Map aid -> internal item token id (string trong vocab)\n ids = [iid2token.get(str(a), None) for a in aids]\n ids = [i for i in ids if i is not None]\n\n if len(ids) == 0:\n ids = [0] # fallback: padding id\n\n ids = ids[-MAX_LEN:]\n lengths.append(len(ids))\n pad_len = MAX_LEN - len(ids)\n seqs.append([0] * pad_len + ids) # padding left\n\n item_id_list = torch.tensor(seqs, dtype=torch.long, device=config[\"device\"])\n item_length = torch.tensor(lengths, dtype=torch.long, device=config[\"device\"])\n\n batch = {\n \"item_id_list\": item_id_list,\n \"item_length\": item_length,\n }\n return batch\n\nTOPK = 20\nBATCH_SIZE_INFER = 512\n\nall_session_preds = {}\n\nfor i in tqdm(range(0, len(test_sessions), BATCH_SIZE_INFER), desc=\"Inference\"):\n batch_sessions = test_sessions[i : i + BATCH_SIZE_INFER]\n session_ids = [s[0] for s in batch_sessions]\n session_aids = [s[1] for s in batch_sessions]\n\n batch_input = build_input_batch(session_aids)\n\n with torch.no_grad():\n scores = model.full_sort_predict(batch_input) # [B, n_items]\n\n top_scores, top_indices = torch.topk(scores, k=TOPK, dim=1)\n top_indices = top_indices.cpu().numpy()\n\n for sid, idx_row in zip(session_ids, top_indices):\n aids_pred = [int(token2iid[int(i)]) for i in idx_row]\n all_session_preds[sid] = aids_pred\n\nlen(all_session_preds)","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"rows = []\n\nfor session_id, preds in all_session_preds.items():\n pred_str = \" \".join(str(a) for a in preds)\n rows.append((f\"{session_id}_clicks\", pred_str))\n rows.append((f\"{session_id}_carts\", pred_str))\n rows.append((f\"{session_id}_orders\", pred_str))\n\nsubmission_df = pd.DataFrame(rows, columns=[\"session_type\", \"labels\"])\nsubmission_df.head()","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"submission_path = Path(\"submission_sample.csv\") if TEST_MODE else Path(\"submission.csv\")\nsubmission_df.to_csv(submission_path, index=False)\nsubmission_path, submission_df.shape","metadata":{"trusted":true},"outputs":[],"execution_count":null}]}