Skip to content

在舊款 CPU 編譯 Tensorflow v2.1

紀錄在舊款 CPU 上編譯 Tensorflow v2.1 的過程，包含安裝 Python 虛擬環境、下載 Tensorflow 源碼、安裝 Bazel、設定環境並編譯等步驟。

參考官網: https://www.tensorflow.org/install/source

硬體版本: OS: Ubuntu 18.04 Server CPU: Intel X5670

軟體版本: Python: Anaconda Python 3.7.5 Bazel: 0.27.1 Tensorflow: r2.1

安裝 Python 虛擬環境

Conda

conda create -n tf2 python=3.7.5
conda activate tf2
conda install pip six numpy wheel setuptools mock
conda install -c conda-forge keras-applications
conda install -c conda-forge keras-preprocessing

Pip

python3 -m venv .venv
source .venv/bin/activate
pip install --upgrade pip setuptools wheel
pip install -U six 'numpy<1.19.0' mock 'future>=0.17.1' 'gast==0.3.3' typing_extensions
pip install -U keras_applications --no-deps
pip install -U keras_preprocessing --no-deps

下載 Tensorflow 源碼並切換 branch

git clone https://github.com/tensorflow/tensorflow.git
cd tensorflow
git checkout r2.1

安裝 Bazel

參考官網對bazel版本限制參考: https://www.tensorflow.org/install/source#tested_build_configurations

要準確知道 tensorflow 限制的 bazel 版本,可參考 tensorflow/configure.py 內的 _TF_MIN_BAZEL_VERSION 和 _TF_MAX_BAZEL_VERSION 2個常數

bazel安裝參考: https://docs.bazel.build/versions/master/install-ubuntu.html

wget https://github.com/bazelbuild/bazel/releases/download/0.27.1/bazel-0.27.1-installer-linux-x86_64.sh
chmod +x bazel-0.27.1-installer-linux-x86_64.sh
./bazel-0.27.1-installer-linux-x86_64.sh --user

更新 .bashrc

PATH=$PATH:$HOME/bin

設定環境並編譯

設定環境

source ~/.bashrc
conda activate tf2

cd tensorflow
bazel clean
$ ./configure # 基本上所有安裝選項都選**N**
Please specify optimization flags to use during compilation when bazel option "--config=opt" is specified [Default is -march=native -Wno-sign-compare]:

Would you like to interactively configure ./WORKSPACE for Android builds? [y/N]: n
Not configuring the WORKSPACE for Android builds.

Preconfigured Bazel build configs. You can use any of the below by adding "--config=<>" to your build command. See .bazelrc for more details.
    --config=mkl                # Build with MKL support.
    --config=monolithic         # Config for mostly static monolithic build.
    --config=ngraph             # Build with Intel nGraph support.
    --config=numa               # Build with NUMA support.
    --config=dynamic_kernels    # (Experimental) Build kernels into separate shared objects.
    --config=v2                 # Build TensorFlow 2.x instead of 1.x.
Preconfigured Bazel build configs to DISABLE default on features:
    --config=noaws              # Disable AWS S3 filesystem support.
    --config=nogcp              # Disable GCP support.
    --config=nohdfs             # Disable HDFS support.
    --config=nonccl             # Disable NVIDIA NCCL support.
Configuration finished

開始編譯

bazel build tensorflow/tools/pip_package:build_pip_package

bazel build --local_cpu_resources=32 --config=opt tensorflow/tools/pip_package:build_pip_package

時間紀錄

CPU: Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz * 2

INFO: Elapsed time: 5665.291s, Critical Path: 4042.33s
INFO: 14283 processes: 14283 local.
INFO: Build completed successfully, 15315 total actions

安裝 tensorflow

用 pip 安裝

./tensorflow/tools/pip_package/build_pip_package.sh /tmp/tensorflow_pkg
pip install /tmp/tensorflow_pkg/tensorflow-2.1.0rc0-cp37-cp37m-linux_x86_64.whl

測試以下代碼

from __future__ import absolute_import, division, print_function, unicode_literals
import tensorflow as tf

mnist = tf.keras.datasets.mnist

(x_train, y_train), (x_test, y_test) = mnist.load_data()
x_train, x_test = x_train / 255.0, x_test / 255.0

model = tf.keras.models.Sequential([
  tf.keras.layers.Flatten(input_shape=(28, 28)),
  tf.keras.layers.Dense(128, activation='relu'),
  tf.keras.layers.Dropout(0.2),
  tf.keras.layers.Dense(10, activation='softmax')
])

model.compile(optimizer='adam',
              loss='sparse_categorical_crossentropy',
              metrics=['accuracy'])

model.fit(x_train, y_train, epochs=5)

model.evaluate(x_test,  y_test, verbose=2)