編譯tensorflow遇見JVM out錯誤

1、問題

[root@k8s-master tensorflow]# bazel build --config=opt --verbose_failures //tensorflow:libtensorflow_cc.so
INFO: Analysed target //tensorflow:libtensorflow_cc.so (138 packages loaded, 11509 targets configured).
INFO: Found 1 target...
[10,192 / 10,400] 48 actions running
    Compiling tensorflow/core/kernels/cwise_op_igammas.cc; 33s local
    Compiling tensorflow/core/kernels/cwise_op_greater.cc; 33s local
    Compiling tensorflow/core/kernels/cwise_op_floor_mod.cc; 33s local
    Compiling tensorflow/core/kernels/cwise_op_mod.cc; 33s local
    Compiling tensorflow/core/kernels/cwise_op_mul_2.cc; 33s local
    Compiling tensorflow/core/kernels/cwise_op_equal_to_2.cc; 33s local
    Compiling tensorflow/core/kernels/cwise_op_pow.cc; 33s local
    Compiling tensorflow/core/kernels/cwise_op_bitwise_and.cc; 33s local ...

Server terminated abruptly (error code: 14, error message: 'Socket closed', log file: '/root/.cache/bazel/_bazel_root/c67f7401a5c0cf4a446e6a7f5e6a0388/server/jvm.out')

2、解決

2.1 查看是否內存問題 即交換內存

free -m 查看下
如果沒有swap內存 增加開啓swap內存、重啓系統

2.2 因爲是用的CUDA 看下GPU的溫度

nvidia-smi 溫度太高 需要進行降溫

3、參考

【1】 編譯tensorflow出現jvm out問題解決