{
    "dataset": "Ali_Display_Ad_Click / Taobao",
    "sample_size": {
        "total": 23601301,
        "train": 17822296,
        "valid": 2871767,
        "test": 2907238,
        "phase1_total_interactions": 26557961
    },
    "split_by_date": {
        "timezone": "Asia/Shanghai",
        "timestamp_unit": "s",
        "train_days": 6,
        "train_range": "2017-05-06 ~ 2017-05-11",
        "valid_days": 1,
        "valid_range": "2017-05-12 ~ 2017-05-12",
        "test_days": 1,
        "test_range": "2017-05-13 ~ 2017-05-13",
        "rule": "按 raw_sample 日期排序后按 train/valid/test 比例切分；官方 8 天数据默认约为 6/1/1。"
    },
    "behavior_log_usage": {
        "used": true,
        "label_columns": [
            "cart",
            "fav",
            "buy"
        ],
        "stats": {
            "raw_rows": 723268134,
            "kept_rows": 33592031,
            "label_counts": {
                "cart": 15943875,
                "fav": 9299893,
                "buy": 8357307
            }
        },
        "match_key": [
            "user_id",
            "cate_id",
            "brand"
        ],
        "behavior_label_window_seconds": 86400,
        "rule": "raw_sample 先通过 adgroup_id 映射到 ad_feature.cate_id/brand；对每条曝光样本，若同一 user_id + cate_id + brand 在曝光 time_stamp 之后 behavior_label_window_seconds 秒内出现对应 btag，则该行为标签置为 1，否则为 0。",
        "note": "behavior_log 不含 adgroup_id，因此这里是类目/品牌粒度、曝光后时间窗口内的行为监督，不是广告 ID 粒度的直接标签。"
    },
    "user_filtering": {
        "min_user_interactions": 10,
        "valid_users": 470570,
        "dropped_users": 671159
    },
    "oov_filter": {
        "min_feat_count": 2,
        "applied_to": [
            "cms_segid",
            "cms_group_id",
            "final_gender_code",
            "age_level",
            "pvalue_level",
            "shopping_level",
            "occupation",
            "new_user_class_level",
            "cate_id",
            "campaign_id",
            "customer_id",
            "brand",
            "price_bucket"
        ],
        "rule": "特征值出现次数 < min_feat_count 的统一映射为 0 (unknown/padding)",
        "freq_source": "基于 raw_sample 中 user/item 出现次数 * 静态特征值统计"
    },
    "blocked_layout": {
        "n_user_parts": 32,
        "train_blocks": 32,
        "valid_blocks": 8,
        "test_blocks": 8,
        "train": {
            "data_dir": "train/data",
            "user_info_dir": "train/user_info",
            "item_info_dir": "train/item_info"
        },
        "valid": {
            "data_dir": "valid/data",
            "user_info_dir": "valid/user_info",
            "item_info_dir": "valid/item_info"
        },
        "test": {
            "data_dir": "test/data",
            "user_info_dir": "test/user_info",
            "item_info_dir": "test/item_info"
        },
        "block_pair_rule": "同一 split 下，data/user_info/item_info 使用相同 part-xxxxx 编号配对读取。",
        "local_index_rule": {
            "user_index": "block-local dense index, starts from 0",
            "item_index": "block-local dense index, 0 reserved for padding",
            "user_id": "global feature id, consistent across blocks",
            "item_id": "global feature id, consistent across blocks"
        }
    },
    "vocab_size": {
        "user_index": 470570,
        "item_index": 846812,
        "user_id": 470571,
        "item_id": 846812,
        "action": 17,
        "timestamp": 0,
        "is_weekend": 3,
        "hour": 25,
        "cms_segid": 99,
        "cms_group_id": 15,
        "final_gender_code": 4,
        "age_level": 9,
        "pvalue_level": 5,
        "shopping_level": 5,
        "occupation": 4,
        "new_user_class_level": 6,
        "cate_id": 6409,
        "campaign_id": 354106,
        "customer_id": 220826,
        "brand": 88145,
        "price_bucket": 11,
        "pid": 4
    },
    "label": [
        "is_click",
        "cart",
        "fav",
        "buy"
    ],
    "action_vocab": {
        "buy": 1,
        "cart": 2,
        "cart|buy": 3,
        "cart|fav": 4,
        "cart|fav|buy": 5,
        "exposure": 6,
        "fav": 7,
        "fav|buy": 8,
        "is_click": 9,
        "is_click|buy": 10,
        "is_click|cart": 11,
        "is_click|cart|buy": 12,
        "is_click|cart|fav": 13,
        "is_click|cart|fav|buy": 14,
        "is_click|fav": 15,
        "is_click|fav|buy": 16
    },
    "action_vocab_desc": "编码后的多任务 action 词表，用于 dataloader 基于 full_action_seq 构造 task-specific token masks。",
    "user_info_schema": {
        "fields": [
            "user_index",
            "full_item_seq",
            "full_action_seq",
            "full_timestamp_seq"
        ],
        "full_timestamp_seq_desc": "按时间顺序排列的 raw_sample time_stamp 序列",
        "desc": "user_index / full_item_seq 中的 item index 为 block-local index；full_action_seq / full_timestamp_seq 为全局时间顺序序列。"
    },
    "item_info_schema": {
        "fields": [
            "item_index",
            "item_id",
            "cate_id",
            "campaign_id",
            "customer_id",
            "brand",
            "price_bucket"
        ],
        "desc": "item_index 为 block-local index；item_id 为全局 item feature id。"
    },
    "feature_schema": {
        "user_static_features": [
            "cms_segid",
            "cms_group_id",
            "final_gender_code",
            "age_level",
            "pvalue_level",
            "shopping_level",
            "occupation",
            "new_user_class_level"
        ],
        "context_features": [
            "pid",
            "is_weekend",
            "hour"
        ],
        "item_static_features": [
            "cate_id",
            "campaign_id",
            "customer_id",
            "brand",
            "price_bucket"
        ],
        "label_columns": [
            "is_click",
            "cart",
            "fav",
            "buy"
        ]
    },
    "max_len": {
        "full_item_seq": 3756,
        "full_action_seq": 3756,
        "full_timestamp_seq": 3756
    }
}