TensorFlow2.6.5配套的Python版本是:Python3.7.x(3.7.5~3.7.11)、Python3.8.x、Python3.9.x。
由于TensorFlow依赖h5py,而h5py依赖HDF5,需要先编译安装HDF5,否则使用pip安装h5py会报错,以下步骤以root用户操作。
tar -zxvf hdf5-1.10.5.tar.gz
cd hdf5-1.10.5/ ./configure --prefix=/usr/local/hdf5 make -j16 && make install
export CPATH=/usr/local/hdf5/include/:/usr/local/hdf5/lib/ export LD_LIBRARY_PATH=/usr/local/hdf5/lib/:$LD_LIBRARY_PATH
pip3 install "Cython<3" pip3 install wheel
pip3 install h5py==3.1.0
若在线安装h5py 3.1.0失败,可单击链接获取源码包,使用源码编译安装:
unzip h5py-3.1.0.zip cd h5py-3.1.0 python3 setup.py build python3 setup.py install
需要安装TensorFlow才可以进行算子开发验证、训练业务开发。
pip3 install tensorflow-cpu==2.6.5
在下载完tensorflow tag v2.6.5源码后需要执行如下步骤:
tf_http_archive( name = "nsync", sha256 = "caf32e6b3d478b78cff6c2ba009c3400f8251f646804bcb65465666a9cea93c4", strip_prefix = "nsync-1.22.0", system_build_file = clean_dep("//third_party/systemlibs:nsync.BUILD"), urls = [ "https://storage.googleapis.com/mirror.tensorflow.org/github.com/google/nsync/archive/1.22.0.tar.gz", "https://github.com/google/nsync/archive/1.22.0.tar.gz", ], )
在NSYNC_CPP_START_内容后添加如下加粗字体内容。
#include "nsync_cpp.h" #include "nsync_atomic.h" NSYNC_CPP_START_ #define ATM_CB_() __sync_synchronize() static INLINE int atm_cas_nomb_u32_ (nsync_atomic_uint32_ *p, uint32_t o, uint32_t n) { int result = (std::atomic_compare_exchange_strong_explicit (NSYNC_ATOMIC_UINT32_PTR_ (p), &o, n, std::memory_order_relaxed, std::memory_order_relaxed)); ATM_CB_(); return result; } static INLINE int atm_cas_acq_u32_ (nsync_atomic_uint32_ *p, uint32_t o, uint32_t n) { int result = (std::atomic_compare_exchange_strong_explicit (NSYNC_ATOMIC_UINT32_PTR_ (p), &o, n, std::memory_order_acquire, std::memory_order_relaxed)); ATM_CB_(); return result; } static INLINE int atm_cas_rel_u32_ (nsync_atomic_uint32_ *p, uint32_t o, uint32_t n) { int result = (std::atomic_compare_exchange_strong_explicit (NSYNC_ATOMIC_UINT32_PTR_ (p), &o, n, std::memory_order_release, std::memory_order_relaxed)); ATM_CB_(); return result; } static INLINE int atm_cas_relacq_u32_ (nsync_atomic_uint32_ *p, uint32_t o, uint32_t n) { int result = (std::atomic_compare_exchange_strong_explicit (NSYNC_ATOMIC_UINT32_PTR_ (p), &o, n, std::memory_order_acq_rel, std::memory_order_relaxed)); ATM_CB_(); return result; }
将上个步骤中解压出的内容压缩为一个新的“nsync-1.22.0.tar.gz”源码包,保存(比如,保存在“/tmp/nsync-1.22.0.tar.gz”)。
sha256sum /tmp/nsync-1.22.0.tar.gz
tf_http_archive( name = "nsync", sha256 = "caf32e6b3d478b78cff6c2ba009c3400f8251f646804bcb65465666a9cea93c4", strip_prefix = "nsync-1.22.0", system_build_file = clean_dep("//third_party/systemlibs:nsync.BUILD"), urls = [ "https://storage.googleapis.com/mirror.tensorflow.org/github.com/google/nsync/archive/1.22.0.tar.gz", "file:///tmp/nsync-1.22.0.tar.gz ", "https://github.com/google/nsync/archive/1.22.0.tar.gz", ], )
执行完./configure之后,需要修改 .tf_configure.bazelrc 配置文件。
添加如下一行build编译选项:
build:opt --cxxopt=-D_GLIBCXX_USE_CXX11_ABI=0
删除以下两行:
build:opt --copt=-march=native build:opt --host_copt=-march=native
以上步骤执行完后会打包TensorFlow到指定目录,进入指定目录后执行如下命令安装:
如下命令如果使用非root用户安装,需要在安装命令后加上--user,例如:pip3 install tensorflow-2.6.5-*.whl --user
pip3 install tensorflow-2.6.5-*.whl
python3 -c "import tensorflow as tf; print(tf.reduce_sum(tf.random.normal([1000, 1000])))"
如果返回了张量则表示安装成功。安装TensorFlow时会自动重装numpy,导入时如果提示numpy版本不兼容,请参考安装TensorFlow2.6.5后,执行import tensorflow时报错重装配套版本的numpy。