Skip to content

  • 体验新版
    • 正在加载...
  • 登录
  • PaddlePaddle
  • Paddle
  • Issue
  • #5670

P
Paddle
  • 项目概览

PaddlePaddle / Paddle
大约 2 年 前同步成功

通知 2325
Star 20933
Fork 5424
  • 代码
    • 文件
    • 提交
    • 分支
    • Tags
    • 贡献者
    • 分支图
    • Diff
  • Issue 1423
    • 列表
    • 看板
    • 标记
    • 里程碑
  • 合并请求 543
  • Wiki 0
    • Wiki
  • 分析
    • 仓库
    • DevOps
  • 项目成员
  • Pages
P
Paddle
  • 项目概览
    • 项目概览
    • 详情
    • 发布
  • 仓库
    • 仓库
    • 文件
    • 提交
    • 分支
    • 标签
    • 贡献者
    • 分支图
    • 比较
  • Issue 1,423
    • Issue 1,423
    • 列表
    • 看板
    • 标记
    • 里程碑
  • 合并请求 543
    • 合并请求 543
  • Pages
  • 分析
    • 分析
    • 仓库分析
    • DevOps
  • Wiki 0
    • Wiki
  • 成员
    • 成员
  • 收起侧边栏
  • 动态
  • 分支图
  • 创建新Issue
  • 提交
  • Issue看板
已关闭
开放中
Opened 11月 15, 2017 by saxon_zh@saxon_zhGuest

mpi训练时cost恒为0

Created by: cszhou

本地训练网络的cost会有变化。但是mpi训练,第一个batch就发现cost为0,接下来cost一直保持不变。请教是什么原因。

目前打印过输入的训练数据,数据看起来没有异常。

I1115 15:30:28.423450  2915 PyDataProvider2.cpp:611] Scanner created.
I1115 15:30:28.468631  2915 PyDataProvider2.cpp:643] Reading CPU Batch Done.
I1115 15:30:44.074748  2915 TrainerInternal.cpp:165]  Batch=1 samples=512 AvgCost=0 CurrentCost=0 Eval: __auc_evaluator_0__=0  CurrentEval: __auc_evaluator_0__=0 
I1115 15:30:44.179998  2915 Stat.cpp:130] ======= StatSet: [GlobalStatInfo] status ======
I1115 15:30:44.180039  2915 Stat.cpp:133] Stat=sendSparseParam                TID=3055   total=12631      avg=12631      max=12631      min=12631      count=1         
I1115 15:30:44.180058  2915 Stat.cpp:133] Stat=clearGradient                  TID=3317   total=1596.19    avg=1596.19    max=1596.19    min=1596.19    count=1         
                                          Stat=clearGradient                  TID=3320   total=2108.17    avg=2108.17    max=2108.17    min=2108.17    count=1         
                                          Stat=clearGradient                  TID=3324   total=2166.38    avg=2166.38    max=2166.38    min=2166.38    count=1         
                                          Stat=clearGradient                  TID=3318   total=2249.74    avg=2249.74    max=2249.74    min=2249.74    count=1         
                                          Stat=clearGradient                  TID=3319   total=2274.27    avg=2274.27    max=2274.27    min=2274.27    count=1         
                                          Stat=clearGradient                  TID=3321   total=2275.64    avg=2275.64    max=2275.64    min=2275.64    count=1         
                                          Stat=clearGradient                  TID=3322   total=2275.71    avg=2275.71    max=2275.71    min=2275.71    count=1         
                                          Stat=clearGradient                  TID=3323   total=2302.63    avg=2302.63    max=2302.63    min=2302.63    count=1         
I1115 15:30:44.180104  2915 Stat.cpp:133] Stat=PyDP2.getNextBatchInternal     TID=2915   total=290085     avg=290085     max=290085     min=290085     count=1         
I1115 15:30:44.180115  2915 Stat.cpp:133] Stat=getParamSparse                 TID=3055   total=43.291     avg=43.291     max=43.291     min=43.291     count=1         
I1115 15:30:44.180127  2915 Stat.cpp:133] Stat=prefetch                       TID=2915   total=64.268     avg=64.268     max=64.268     min=64.268     count=1         
I1115 15:30:44.180140  2915 Stat.cpp:133] Stat=wait_value                     TID=3317   total=0          avg=0          max=0          min=0          count=1         
                                          Stat=wait_value                     TID=3320   total=0.001      avg=0.001      max=0.001      min=0.001      count=1         
                                          Stat=wait_value                     TID=3324   total=0          avg=0          max=0          min=0          count=1         
                                          Stat=wait_value                     TID=3318   total=0.001      avg=0.001      max=0.001      min=0.001      count=1         
                                          Stat=wait_value                     TID=3319   total=0          avg=0          max=0          min=0          count=1         
                                          Stat=wait_value                     TID=3321   total=0.001      avg=0.001      max=0.001      min=0.001      count=1         
                                          Stat=wait_value                     TID=3322   total=0          avg=0          max=0          min=0          count=1         
                                          Stat=wait_value                     TID=3323   total=0          avg=0          max=0          min=0          count=1         
I1115 15:30:44.180238  2915 Stat.cpp:133] Stat=TrainBatch                     TID=2915   total=15711.2    avg=15711.2    max=15711.2    min=15711.2    count=1         
I1115 15:30:44.180255  2915 Stat.cpp:133] Stat=waitOutArgs                    TID=2915   total=2302.62    avg=287.828    max=1596.21    min=0.001      count=8         
I1115 15:30:44.180269  2915 Stat.cpp:133] Stat=getTrainBatch                  TID=2915   total=290085     avg=290085     max=290085     min=290085     count=1         
I1115 15:30:44.180279  2915 Stat.cpp:133] Stat=forwardBackward                TID=2915   total=2759.93    avg=2759.93    max=2759.93    min=2759.93    count=1         
I1115 15:30:44.180292  2915 Stat.cpp:133] Stat=finishBatch                    TID=2915   total=12631.1    avg=12631.1    max=12631.1    min=12631.1    count=1         
I1115 15:30:44.180305  2915 Stat.cpp:133] Stat=client_sendAndRecv_send        TID=7585   total=67.639     avg=33.819     max=36.404     min=31.235     count=2         
                                          Stat=client_sendAndRecv_send        TID=7601   total=309.091    avg=154.545    max=179.962    min=129.129    count=2         
                                          Stat=client_sendAndRecv_send        TID=7599   total=285.496    avg=142.748    max=151.026    min=134.47     count=2         
                                          Stat=client_sendAndRecv_send        TID=7594   total=313.316    avg=156.658    max=161.559    min=151.757    count=2         
                                          Stat=client_sendAndRecv_send        TID=7595   total=318.813    avg=159.406    max=159.71     min=159.103    count=2         
                                          Stat=client_sendAndRecv_send        TID=7586   total=324.149    avg=162.074    max=162.552    min=161.597    count=2         
                                          Stat=client_sendAndRecv_send        TID=7587   total=316.404    avg=158.202    max=162.621    min=153.783    count=2         
                                          Stat=client_sendAndRecv_send        TID=7591   total=507.3      avg=253.65     max=344.283    min=163.017    count=2         
                                          Stat=client_sendAndRecv_send        TID=7593   total=325.368    avg=162.684    max=164.66     min=160.708    count=2         
                                          Stat=client_sendAndRecv_send        TID=7603   total=324.085    avg=162.042    max=164.879    min=159.206    count=2         
                                          Stat=client_sendAndRecv_send        TID=7596   total=450.1      avg=225.05     max=283.793    min=166.307    count=2         
                                          Stat=client_sendAndRecv_send        TID=7602   total=331.816    avg=165.908    max=167.097    min=164.719    count=2         
                                          Stat=client_sendAndRecv_send        TID=7590   total=330.372    avg=165.186    max=167.269    min=163.103    count=2         
                                          Stat=client_sendAndRecv_send        TID=7588   total=331.817    avg=165.908    max=167.757    min=164.06     count=2         
                                          Stat=client_sendAndRecv_send        TID=7592   total=331.824    avg=165.912    max=167.808    min=164.016    count=2         
                                          Stat=client_sendAndRecv_send        TID=7604   total=330.974    avg=165.487    max=168.348    min=162.626    count=2         
                                          Stat=client_sendAndRecv_send        TID=7597   total=470.18     avg=235.09     max=284.13     min=186.05     count=2         
                                          Stat=client_sendAndRecv_send        TID=7598   total=377.881    avg=188.94     max=214.039    min=163.842    count=2         
                                          Stat=client_sendAndRecv_send        TID=7600   total=506.221    avg=253.11     max=291.278    min=214.943    count=2         
                                          Stat=client_sendAndRecv_send        TID=7589   total=466.909    avg=233.454    max=305.997    min=160.912    count=2         
                                          Stat=client_sendAndRecv_send        TID=7574   total=2546.58    avg=848.859    max=2545.53    min=0.397      count=3         
                                          Stat=client_sendAndRecv_send        TID=7578   total=2666.93    avg=888.978    max=2665.68    min=0.431      count=3         
                                          Stat=client_sendAndRecv_send        TID=7576   total=2697.64    avg=899.213    max=2696.71    min=0.169      count=3         
                                          Stat=client_sendAndRecv_send        TID=7563   total=2706.02    avg=902.006    max=2705.02    min=0.139      count=3         
                                          Stat=client_sendAndRecv_send        TID=7568   total=2936.91    avg=978.97     max=2935.77    min=0.474      count=3         
                                          Stat=client_sendAndRecv_send        TID=7573   total=3083.46    avg=1027.82    max=3082.01    min=0.554      count=3         
                                          Stat=client_sendAndRecv_send        TID=7564   total=3291.76    avg=1097.25    max=3290.58    min=0.307      count=3         
                                          Stat=client_sendAndRecv_send        TID=7570   total=3350.56    avg=1116.85    max=3348.91    min=0.278      count=3         
                                          Stat=client_sendAndRecv_send        TID=7580   total=3430.14    avg=1143.38    max=3428.91    min=0.425      count=3         
                                          Stat=client_sendAndRecv_send        TID=7572   total=3482.64    avg=1160.88    max=3481.21    min=0.455      count=3         
                                          Stat=client_sendAndRecv_send        TID=7582   total=3530.81    avg=1176.94    max=3530.14    min=0.296      count=3         
                                          Stat=client_sendAndRecv_send        TID=7577   total=3531.34    avg=1177.11    max=3530.16    min=0.237      count=3         
                                          Stat=client_sendAndRecv_send        TID=7566   total=3574.35    avg=1191.45    max=3573.46    min=0.37       count=3         
                                          Stat=client_sendAndRecv_send        TID=7569   total=3585.07    avg=1195.02    max=3583.98    min=0.247      count=3         
                                          Stat=client_sendAndRecv_send        TID=7575   total=3588.39    avg=1196.13    max=3587.25    min=0.444      count=3         
                                          Stat=client_sendAndRecv_send        TID=7579   total=3587.49    avg=1195.83    max=3585.95    min=0.504      count=3         
                                          Stat=client_sendAndRecv_send        TID=7571   total=3599.68    avg=1199.89    max=3598.16    min=0.529      count=3         
                                          Stat=client_sendAndRecv_send        TID=7565   total=3608.49    avg=1202.83    max=3607.17    min=0.527      count=3         
                                          Stat=client_sendAndRecv_send        TID=7567   total=3702.2     avg=1234.07    max=3700.89    min=0.584      count=3         
                                          Stat=client_sendAndRecv_send        TID=7581   total=3962.36    avg=1320.79    max=3961.32    min=0.336      count=3         
I1115 15:30:44.180513  2915 Stat.cpp:133] Stat=client_sendAndRecv_recv        TID=7585   total=11996      avg=5997.99    max=11913.9    min=82.038     count=2         
                                          Stat=client_sendAndRecv_recv        TID=7601   total=11818.5    avg=5909.26    max=11597.7    min=220.791    count=2         
                                          Stat=client_sendAndRecv_recv        TID=7599   total=12263.1    avg=6131.56    max=12038.9    min=224.22     count=2         
                                          Stat=client_sendAndRecv_recv        TID=7594   total=11692.8    avg=5846.42    max=11484.4    min=208.413    count=2         
                                          Stat=client_sendAndRecv_recv        TID=7597   total=12123.5    avg=6061.74    max=11947.3    min=176.218    count=2         
                                          Stat=client_sendAndRecv_recv        TID=7591   total=12033.3    avg=6016.64    max=11827.5    min=205.81     count=2         
                                          Stat=client_sendAndRecv_recv        TID=7592   total=11738.3    avg=5869.15    max=11536.7    min=201.602    count=2         
                                          Stat=client_sendAndRecv_recv        TID=7587   total=12260.7    avg=6130.36    max=12051.2    min=209.478    count=2         
                                          Stat=client_sendAndRecv_recv        TID=7603   total=12128.9    avg=6064.47    max=11921.7    min=207.265    count=2         
                                          Stat=client_sendAndRecv_recv        TID=7595   total=11881.4    avg=5940.72    max=11666.8    min=214.593    count=2         
                                          Stat=client_sendAndRecv_recv        TID=7586   total=12319.9    avg=6159.95    max=12106.5    min=213.362    count=2         
                                          Stat=client_sendAndRecv_recv        TID=7593   total=12187.6    avg=6093.82    max=11975.7    min=211.971    count=2         
                                          Stat=client_sendAndRecv_recv        TID=7596   total=11851.8    avg=5925.88    max=11641      min=210.795    count=2         
                                          Stat=client_sendAndRecv_recv        TID=7604   total=12302.4    avg=6151.22    max=12091.2    min=211.289    count=2         
                                          Stat=client_sendAndRecv_recv        TID=7590   total=12231.3    avg=6115.65    max=12018.4    min=212.955    count=2         
                                          Stat=client_sendAndRecv_recv        TID=7588   total=11727.3    avg=5863.64    max=11513.6    min=213.711    count=2         
                                          Stat=client_sendAndRecv_recv        TID=7598   total=11694.9    avg=5847.47    max=11526.5    min=168.474    count=2         
                                          Stat=client_sendAndRecv_recv        TID=7600   total=11113.1    avg=5556.53    max=10944.3    min=168.743    count=2         
                                          Stat=client_sendAndRecv_recv        TID=7602   total=11899.5    avg=5949.77    max=11680.3    min=219.22     count=2         
                                          Stat=client_sendAndRecv_recv        TID=7589   total=11802.3    avg=5901.15    max=11712.1    min=90.198     count=2         
                                          Stat=client_sendAndRecv_recv        TID=7574   total=22538.9    avg=7512.96    max=11683.7    min=33.286     count=3         
                                          Stat=client_sendAndRecv_recv        TID=7578   total=22555.6    avg=7518.53    max=11733.4    min=32.712     count=3         
                                          Stat=client_sendAndRecv_recv        TID=7576   total=22933.2    avg=7644.4     max=11750      min=33.023     count=3         
                                          Stat=client_sendAndRecv_recv        TID=7563   total=22975.5    avg=7658.51    max=11712.8    min=33.845     count=3         
                                          Stat=client_sendAndRecv_recv        TID=7568   total=22831.3    avg=7610.43    max=11697.4    min=33.393     count=3         
                                          Stat=client_sendAndRecv_recv        TID=7573   total=22844.6    avg=7614.85    max=11761.6    min=33.88      count=3         
                                          Stat=client_sendAndRecv_recv        TID=7564   total=22854.2    avg=7618.07    max=11683.3    min=34.179     count=3         
                                          Stat=client_sendAndRecv_recv        TID=7570   total=22966.5    avg=7655.52    max=11810.9    min=34.095     count=3         
                                          Stat=client_sendAndRecv_recv        TID=7569   total=22662.7    avg=7554.23    max=11695.4    min=33.218     count=3         
                                          Stat=client_sendAndRecv_recv        TID=7579   total=22691.3    avg=7563.75    max=11711.9    min=33.803     count=3         
                                          Stat=client_sendAndRecv_recv        TID=7566   total=22762.1    avg=7587.36    max=11758.3    min=33.807     count=3         
                                          Stat=client_sendAndRecv_recv        TID=7580   total=22926.2    avg=7642.08    max=11739.1    min=34.922     count=3         
                                          Stat=client_sendAndRecv_recv        TID=7577   total=22837.7    avg=7612.57    max=11732.2    min=34.615     count=3         
                                          Stat=client_sendAndRecv_recv        TID=7571   total=22755.9    avg=7585.28    max=11694.8    min=31.967     count=3         
                                          Stat=client_sendAndRecv_recv        TID=7582   total=22840.8    avg=7613.6     max=11706      min=37.472     count=3         
                                          Stat=client_sendAndRecv_recv        TID=7567   total=22991.1    avg=7663.72    max=11911.1    min=33.903     count=3         
                                          Stat=client_sendAndRecv_recv        TID=7565   total=23024.7    avg=7674.9     max=11675.4    min=33.856     count=3         
                                          Stat=client_sendAndRecv_recv        TID=7575   total=23982.2    avg=7994.05    max=12599.1    min=35.064     count=3         
                                          Stat=client_sendAndRecv_recv        TID=7581   total=22787.3    avg=7595.76    max=11744.6    min=34.744     count=3         
                                          Stat=client_sendAndRecv_recv        TID=7572   total=23539.3    avg=7846.42    max=11848.6    min=33.681     count=3         
I1115 15:30:44.180706  2915 Stat.cpp:133] Stat=waitbeforeMerge                TID=3317   total=823.051    avg=411.525    max=706.432    min=116.619    count=2         
                                          Stat=waitbeforeMerge                TID=3320   total=233.846    avg=116.923    max=194.483    min=39.363     count=2         
                                          Stat=waitbeforeMerge                TID=3324   total=160.284    avg=80.142     max=136.256    min=24.028     count=2         
                                          Stat=waitbeforeMerge                TID=3318   total=52.931     avg=26.465     max=52.91      min=0.021      count=2         
                                          Stat=waitbeforeMerge                TID=3321   total=28.257     avg=14.128     max=27.011     min=1.246      count=2         
                                          Stat=waitbeforeMerge                TID=3322   total=27.629     avg=13.814     max=26.952     min=0.677      count=2         
                                          Stat=waitbeforeMerge                TID=3319   total=82.84      avg=41.42      max=54.453     min=28.387     count=2         
                                          Stat=waitbeforeMerge                TID=3323   total=5.656      avg=2.828      max=5.634      min=0.022      count=2         
I1115 15:30:44.180765  2915 Stat.cpp:133] Stat=copyOutArgs                    TID=2915   total=0.001      avg=0.001      max=0.001      min=0.001      count=1         
I1115 15:30:44.180779  2915 Stat.cpp:133] Stat=taskSem_wait                   TID=3317   total=384087     avg=192043     max=384025     min=61.968     count=2         
                                          Stat=taskSem_wait                   TID=3318   total=384087     avg=192043     max=384025     min=61.972     count=2         
                                          Stat=taskSem_wait                   TID=3319   total=384087     avg=192043     max=384025     min=61.982     count=2         
                                          Stat=taskSem_wait                   TID=3320   total=384087     avg=192043     max=384025     min=61.98      count=2         
                                          Stat=taskSem_wait                   TID=3321   total=384087     avg=192043     max=384025     min=61.985     count=2         
                                          Stat=taskSem_wait                   TID=3322   total=384087     avg=192043     max=384025     min=61.992     count=2         
                                          Stat=taskSem_wait                   TID=3324   total=384087     avg=192043     max=384025     min=62         count=2         
                                          Stat=taskSem_wait                   TID=3323   total=384087     avg=192043     max=384025     min=61.993     count=2         
I1115 15:30:44.180820  2915 Stat.cpp:133] Stat=thread_forward                 TID=3317   total=0          avg=0          max=0          min=0          count=1         
                                          Stat=thread_forward                 TID=3320   total=0          avg=0          max=0          min=0          count=1         
                                          Stat=thread_forward                 TID=3324   total=0          avg=0          max=0          min=0          count=1         
                                          Stat=thread_forward                 TID=3318   total=0          avg=0          max=0          min=0          count=1         
                                          Stat=thread_forward                 TID=3319   total=0          avg=0          max=0          min=0          count=1         
                                          Stat=thread_forward                 TID=3321   total=0.001      avg=0.001      max=0.001      min=0.001      count=1         
                                          Stat=thread_forward                 TID=3322   total=0          avg=0          max=0          min=0          count=1         
                                          Stat=thread_forward                 TID=3323   total=0          avg=0          max=0          min=0          count=1         
I1115 15:30:44.180851  2915 Stat.cpp:133] Stat=thread_backward                TID=3317   total=1163.64    avg=1163.64    max=1163.64    min=1163.64    count=1         
                                          Stat=thread_backward                TID=3320   total=651.695    avg=651.695    max=651.695    min=651.695    count=1         
                                          Stat=thread_backward                TID=3324   total=593.467    avg=593.467    max=593.467    min=593.467    count=1         
                                          Stat=thread_backward                TID=3322   total=484.157    avg=484.157    max=484.157    min=484.157    count=1         
                                          Stat=thread_backward                TID=3321   total=484.222    avg=484.222    max=484.222    min=484.222    count=1         
                                          Stat=thread_backward                TID=3318   total=510.124    avg=510.124    max=510.124    min=510.124    count=1         
                                          Stat=thread_backward                TID=3323   total=457.241    avg=457.241    max=457.241    min=457.241    count=1         
                                          Stat=thread_backward                TID=3319   total=485.59     avg=485.59     max=485.59     min=485.59     count=1         
I1115 15:30:44.180907  2915 Stat.cpp:133] Stat=mergeGradDense                 TID=3317   total=339.396    avg=2.495      max=134.603    min=0.001      count=136       
                                          Stat=mergeGradDense                 TID=3321   total=454.859    avg=3.344      max=248.478    min=0.001      count=136       
                                          Stat=mergeGradDense                 TID=3323   total=450.469    avg=3.312      max=249.04     min=0.001      count=136       
                                          Stat=mergeGradDense                 TID=3322   total=455.439    avg=3.348      max=250.046    min=0          count=136       
                                          Stat=mergeGradDense                 TID=3318   total=456.098    avg=3.353      max=251.478    min=0          count=136       
                                          Stat=mergeGradDense                 TID=3319   total=401.6      avg=2.952      max=211.521    min=0          count=136       
                                          Stat=mergeGradDense                 TID=3320   total=416.759    avg=3.064      max=231.794    min=0.001      count=136       
                                          Stat=mergeGradDense                 TID=3324   total=432.113    avg=3.177      max=239.906    min=0.001      count=136       
I1115 15:30:44.180948  2915 Stat.cpp:133] Stat=eval                           TID=2915   total=0.003      avg=0.003      max=0.003      min=0.003      count=1         
I1115 15:30:44.180960  2915 Stat.cpp:133] Stat=mergeRemoteGradSparse          TID=3321   total=1.029      avg=0.033      max=0.37       min=0.004      count=31        
                                          Stat=mergeRemoteGradSparse          TID=3322   total=1.025      avg=0.033      max=0.394      min=0.001      count=31        
                                          Stat=mergeRemoteGradSparse          TID=3319   total=1.018      avg=0.032      max=0.369      min=0.001      count=31        
                                          Stat=mergeRemoteGradSparse          TID=3317   total=1.028      avg=0.033      max=0.369      min=0.003      count=31        
                                          Stat=mergeRemoteGradSparse          TID=3324   total=0.965      avg=0.031      max=0.369      min=0          count=31        
                                          Stat=mergeRemoteGradSparse          TID=3318   total=1.025      avg=0.033      max=0.39       min=0.001      count=31        
                                          Stat=mergeRemoteGradSparse          TID=3323   total=1.018      avg=0.032      max=0.369      min=0.003      count=31        
                                          Stat=mergeRemoteGradSparse          TID=3320   total=0.978      avg=0.031      max=0.37       min=0.001      count=31        
I1115 15:30:44.180997  2915 Stat.cpp:133] Stat=sumCost                        TID=2915   total=0          avg=0          max=0          min=0          count=1         
I1115 15:30:44.181008  2915 Stat.cpp:133] Stat=sendAndRecv_dense              TID=2915   total=12268.4    avg=12268.4    max=12268.4    min=12268.4    count=1         
I1115 15:30:44.181022  2915 Stat.cpp:157] --------------------------------------------------
指派人
分配到
无
里程碑
无
分配里程碑
工时统计
无
截止日期
无
标识: paddlepaddle/Paddle#5670
渝ICP备2023009037号

京公网安备11010502055752号

网络110报警服务 Powered by GitLab CE v13.7
开源知识
Git 入门 Pro Git 电子书 在线学 Git
Markdown 基础入门 IT 技术知识开源图谱
帮助
使用手册 反馈建议 博客
《GitCode 隐私声明》 《GitCode 服务条款》 关于GitCode
Powered by GitLab CE v13.7