安裝環(huán)境

• OS Platform and Distribution

macos Ventura 13.1

• Architecture

Intel x86_64

• Tensorflow Version

tf 2.11.0

• Python version

3.9.13

• Bazel version

5.3.0

• Xcode version

14.2

從源代碼構(gòu)建

macOS設(shè)置

• miniconda創(chuàng)建虛擬環(huán)境及tensorflow依賴項(xiàng)安裝

conda create -n tf python==3.9.13
conda activate tf
pip install -U pip numpy wheel
pip install -U keras_preprocessing --no-deps

• 安裝Bazel

brew install bazel

• 下載 TensorFlow 源代碼

git clone https://github.com/tensorflow/tensorflow.git
cd tensorflow
git checkout r2.11

配置 build

# Check whether script is executing in a VirtualEnv or Conda environment
if [ -z "$VIRTUAL_ENV" ] && [ -z "$CONDA_PREFIX" ] ; then
	echo "VirtualEnv or Conda env is not activated"
	exit -1
fi

# Set the virtual environment path
if ! [ -z "$VIRTUAL_ENV" ] ; then
  VENV_PATH=$VIRTUAL_ENV
elif ! [ -z "$CONDA_PREFIX" ] ; then
  VENV_PATH=$CONDA_PREFIX
fi

# Set the bin and lib directories
VENV_BIN=$VENV_PATH/bin
VENV_LIB=$VENV_PATH/lib

# bazel tf needs these env vars
export PYTHON_BIN_PATH=$VENV_BIN/python
export PYTHON_LIB_PATH=`ls -d $VENV_LIB/*/ | grep python`

# Set the native architecture optimization flag, which is a default
COPT="--copt=-march=native"

# Determine the available features of your CPU
raw_cpu_flags=`sysctl -a | grep machdep.cpu.features | cut -d ":" -f 2 | tr '[:upper:]' '[:lower:]'`

# Append each of your CPU's features to the list of optimization flags
for cpu_feature in $raw_cpu_flags
do
	case "$cpu_feature" in
		"sse4.1" | "sse4.2" | "ssse3" | "fma" | "cx16" | "popcnt" | "maes")
		    COPT+=" --copt=-m$cpu_feature"
		;;
		"avx1.0")
		    COPT+=" --copt=-mavx"
		;;
		*)
			# noop
		;;
	esac
done

echo $COPT

# First ensure a clear working directory in case you've run bazel previously
bazel clean --expunge

# Run TensorFlow configuration (accept defaults unless you have a need)
python configure.py

# Build the TensorFlow pip package
bazel build -c opt $COPT -k //tensorflow/tools/pip_package:build_pip_package
bazel-bin/tensorflow/tools/pip_package/build_pip_package /tmp/tensorflow_pkg

安裝測(cè)試軟件包

cd ../
pip uninstall tensorflow  # remove current version

pip install /tmp/tensorflow_pkg/tensorflow-version-tags.whl
python -c "import tensorflow as tf; print(tf.__version__)"

編譯異常處理

1、第三方源碼包無(wú)法下載

• 問(wèn)題現(xiàn)象：

	Download from https://storage.googleapis.com/mirror.tensorflow.org/github.com/google/XNNPACK/archive/a50369c0fdd15f0f35b1a91c964644327a88d480.zip failed: class java.io.IOException connect timed out

• 解決方法：

  1. 從該鏈接直接下載文件包，https://github.com/google/XNNPACK/archive/a50369c0fdd15f0f35b1a91c964644327a88d480.zip，（國(guó)內(nèi)下載不了，方法不再贅述）。
  2. 本地搭建http服務(wù)器，用于存放無(wú)法下載的程序包，提供本地下載地址
  3. 修改bzl文件，增加本地鏈接，例如以上鏈接是在tensorflow\tensorflow\workspace2.bzl中

原文內(nèi)容：

tf_http_archive(
    name = "XNNPACK",
    sha256 = "7a16ab0d767d9f8819973dbea1dc45e4e08236f89ab702d96f389fdc78c5855c",
    strip_prefix = "XNNPACK-e8f74a9763aa36559980a0c2f37f587794995622",
    urls = tf_mirror_urls("https://github.com/google/XNNPACK/archive/e8f74a9763aa36559980a0c2f37f587794995622.zip"),
)

替換為：

tf_http_archive(
    name = "XNNPACK",
    sha256 = "ca3a5316b8161214f8f22a578fb638f1fccd0585eee40301363ffd026310379a",
    strip_prefix = "XNNPACK-a50369c0fdd15f0f35b1a91c964644327a88d480",
    urls = ["https://storage.googleapis.com/mirror.tensorflow.org/github.com/google/XNNPACK/archive/a50369c0fdd15f0f35b1a91c964644327a88d480.zip",
            "http://127.0.0.1:8080/tensorflow/XNNPACK-a50369c0fdd15f0f35b1a91c964644327a88d480.zip"
    ],
)

2、Go無(wú)法下載安裝

• 問(wèn)題現(xiàn)象：

  			Download from https://golang.org/dl/?mode=json&include=all failed: class java.io.IOException connect timed out
  

• 解決方法：

1. 該錯(cuò)誤io_bazel_rules_go需要下載go軟件包引起的，需要修改io_bazel_rules_go這個(gè)包中的sdk.bzl文件。從tensorflow\tensorflow\workspace2.bzl中找到這個(gè)包的下載地址，下載到本地，放到本地http服務(wù)器
2. 修改tensorflow\tensorflow\workspace2.bzl文件，注意sha256也需要修改，可以通過(guò)tensorflow編譯錯(cuò)誤獲取，同時(shí)需要?jiǎng)h除/var/tmp/bazel[loginName]/cache/repos/v1/content_addressable/sha256下面和sha256同名文件，避免無(wú)法使用修改后的文件：

原文內(nèi)容：

	tf_http_archive(
    name = "io_bazel_rules_go",
    sha256 = "80cbfe287bacbe6cec97f0446a413fee7c21dafda3981b9de71a35eebbc89e1f",
    urls = tf_mirror_urls("https://github.com/bazelbuild/rules_go/releases/download/v0.34.0/rules_go-v0.34.0.zip"),
    ],
)



	tf_http_archive(
    name = "io_bazel_rules_go",
    sha256 = "80cbfe287bacbe6cec97f0446a413fee7c21dafda3981b9de71a35eebbc89e1f",
    #urls = tf_mirror_urls("https://github.com/bazelbuild/rules_go/releases/download/v0.34.0/rules_go-v0.34.0.zip"),
    urls = ["http://127.0.0.1:8080/tensorflow/rules_go-v0.34.0.zip"
    ],
)

2、下載go.json文件，放到本地http服務(wù)器，sha256值需要計(jì)算傳入，可以通過(guò)替換bzl中某個(gè)包，讓tensorflow編譯時(shí)報(bào)錯(cuò)，提示出該文件的sha256值

3、解壓下載的zip包，打開(kāi) rules*_go-v0.34.0\go\private\sdk.bzl，修改以下內(nèi)容*

原文內(nèi)容：

ctx.download(
            url = [
                "https://golang.org/dl/?mode=json&include=all",
                "https://golang.google.cn/dl/?mode=json&include=all",
            ],
            output = "versions.json",
        )

......

_go_download_sdk = repository_rule(
    implementation = _go_download_sdk_impl,
    attrs = {
        "goos": attr.string(),
        "goarch": attr.string(),
        "sdks": attr.string_list_dict(),
        "urls": attr.string_list(default = ["https://dl.google.com/go/{}"]),
        "version": attr.string(),
        "strip_prefix": attr.string(default = "go"),
    },
)

替換為

ctx.download(
url = [
#"https://golang.org/dl/?mode=json&include=all",
#"https://golang.google.cn/dl/?mode=json&include=all",
"http://127.0.0.1:8080/tensorflow/go.json",
],
output = "versions.json",
sha256 = "47bd74d7eac125a61194d1aa2688c2da3e3bc1bb0981d41a7dd0e9f9cc88d57c",
)

......

_go_download_sdk = repository_rule(
    implementation = _go_download_sdk_impl,
    attrs = {
        "goos": attr.string(),
        "goarch": attr.string(),
        "sdks": attr.string_list_dict(),
        "urls": attr.string_list(default = ["http://127.0.0.1:8080/tensorflow/{}"]),
        "version": attr.string(),
        "strip_prefix": attr.string(default = "go"),
    },
)

4、重新zip壓縮rules_go-v0.34.0.zip，需要根據(jù)編譯錯(cuò)誤提示更改sha256

3、ld: malformed trie, node past end file ‘bazel-out/host/bin/_solib_darwin_x86_64/libtensorflow_Spython_S_Upywrap_Utensorflow_Uinternal.so’

1、該問(wèn)題是由于xcode14升級(jí)引起的，將ld切換成13版本可以暫時(shí)解決該問(wèn)題，查看該鏈接

2、從蘋(píng)果官網(wǎng)下載xcode13，解壓后，將ld拷貝到本機(jī)目錄下，本機(jī)的備份

4、安裝打包好的wheel時(shí)，提示：error:tensorflow-*.whl is not suported wheel on this platform

1、tensorflow, 版本不對(duì)，目前編譯打包成功的時(shí)v2.11.0版本

2、python版本不對(duì)，必須用編譯打包時(shí)使用的python pip安裝

5、dlopen(/Users/davidlaxer/tensorflow-metal/lib/python3.8/site-packages/tensorflow-plugins/libmetal_plugin.dylib, 6): Symbol not found:

1、本地原因是之前安裝了tensorflow-macos，需要卸載之前安裝的所有tensorflow依賴包文章來(lái)源地址http://www.zghlxwxcb.cn/news/detail-861185.html

到了這里，關(guān)于MacOS 系統(tǒng)成功安裝 tensorflow 步驟的文章就介紹完了。如果您還想了解更多內(nèi)容，請(qǐng)?jiān)谟疑辖撬阉鱐OY模板網(wǎng)以前的文章或繼續(xù)瀏覽下面的相關(guān)文章，希望大家以后多多支持TOY模板網(wǎng)！