Created by: ziyenano

线上无法正常预测, 但线下可以正常预测 预测脚本为

paddle cluster_train \
    --config=$cfg \
    --trainer_count=12 \
    --log_period=100 \
    --dot_period=10 \
    --num_passes=100 \
    --use_gpu=cpu \
    --test_period=0 \
    --show_parameter_stats_period=100 \
    --time_limit=700:00:00 \
    --where=nmg01-hpc-off-dmop-slow-cpu-10G_cluster \
    --submitter=caoqixiu \
    --num_nodes=30 \
    --job=test \
    --job_priority=normal \
    --job_name=caoqixiu_ecom_price_predict \
    --thirdparty=./thirdparty \
    2>&1 | tee 'predict.log'

ERROR 如下:

Job Status Series:
epilogue finished
epilogue started
user_main failed (Killed)
job will be killed because agent(16) lost
job will be killed because agent(23) lost
job will be killed because agent(19) lost
user_main started
prologue finished
prologue started
Ready