BQML报错：无法同时通过多创建模型任务更新模型求助

问题描述 #

执行单个BQML模型创建语句时，反复收到「无法同时运行多个模型任务」错误，确认无其他并行模型任务在运行。

涉事代码 #

CREATE OR REPLACE MODEL `analytics-mkt-cleanroom.MKT_DS.HS_KITCHEN_PROEJECT_SELLING_MODEL`
OPTIONS (model_type='BOOSTED_TREE_CLASSIFIER',
input_label_cols=['Y'],
MAX_TREE_DEPTH = 6,
LEARN_RATE = 0.01,
COLSAMPLE_BYTREE = 0.8,
SUBSAMPLE = 0.8,
BOOSTER_TYPE = 'gbtree',
DATA_SPLIT_METHOD = 'RANDOM',
DATA_SPLIT_EVAL_FRACTION = 0.2,
MAX_ITERATIONS = 20,
MAX_PARALLEL_TRIALS = 5,
ENABLE_GLOBAL_EXPLAIN = TRUE,
AUTO_CLASS_WEIGHTS = TRUE
) AS 
SELECT * EXCEPT(HH_ID,id) FROM `analytics-mkt-cleanroom.MKT_DS.KITCHEN_PROJECT_SELLING_TRAIN`

报错截图 #

排查与解决方案 #

1. 清理残留任务或资源锁 #

BigQuery可能存在未正常释放的模型作业锁，即使控制台无运行中任务：

• 进入BigQuery控制台的作业历史，筛选所有状态为「运行中」「排队中」的模型相关作业，手动取消。
• 若作业历史无异常，尝试修改模型名称（如添加_v1后缀），规避原模型资源的隐性冲突。

2. 调整参数配置 #

代码中MAX_PARALLEL_TRIALS = 5用于超参数调优的并行试验，若无需超参调优，建议移除该参数，避免触发多任务逻辑：

CREATE OR REPLACE MODEL `analytics-mkt-cleanroom.MKT_DS.HS_KITCHEN_PROEJECT_SELLING_MODEL`
OPTIONS (model_type='BOOSTED_TREE_CLASSIFIER',
input_label_cols=['Y'],
MAX_TREE_DEPTH = 6,
LEARN_RATE = 0.01,
COLSAMPLE_BYTREE = 0.8,
SUBSAMPLE = 0.8,
BOOSTER_TYPE = 'gbtree',
DATA_SPLIT_METHOD = 'RANDOM',
DATA_SPLIT_EVAL_FRACTION = 0.2,
MAX_ITERATIONS = 20,
-- MAX_PARALLEL_TRIALS = 5, 注释或移除该参数
ENABLE_GLOBAL_EXPLAIN = TRUE,
AUTO_CLASS_WEIGHTS = TRUE
) AS 
SELECT * EXCEPT(HH_ID,id) FROM `analytics-mkt-cleanroom.MKT_DS.KITCHEN_PROJECT_SELLING_TRAIN`

3. 检查项目配额 #

确认项目是否达到模型训练并行任务配额上限：

• 进入BigQuery的IAM与管理 > 配额页面，搜索「模型训练」相关配额，查看使用情况。若配额不足，提交配额提升申请。

内容的提问来源于stack exchange，提问作者Piyush