取消打印Tensorflow中的无用信息，如tensorflow:AutoGraph could not transform ＜*＞ and will run it as-is、加载CUDA信息等

取消打印Tensorflow中的无用信息，如WARNING:tensorflow:AutoGraph could not transform bound method BatchNorm.call of model.BatchNorm object at 0x0000021764BB9F60 and will run it as-is、加载CUDA信息等

• 1.软件环境
• 2.问题描述
  *
• 2.1.CUDA信息以及CPU指令集不支持AVX2
• 2.2.Tensorflow 1.0和Tensorflow 2.0之间的API变更等Warning
• 2.3.Pyinstaller打包为exe后额外增加的Warning
• 3.解决方法
  *
• 3.1.取消打印CUDA信息以及CPU指令集不支持AVX2
  –
  • 3.1.1.具体操作
  • 3.1.2.效果预览
• 3.2.取消打印Tensorflow 2.0API变更以及Pyinstaller打包为exe后额外增加的Warning
  –
  • 3.2.1.具体操作
  • 3.2.2.效果预览
• 4.总结

TensorFlow其他警告信息如何取消：
Python：设置不显示Using TensorFlow backend及FutureWarning: Passing type, 1 or ‘1type’ as a synonym of type is deprecated; in a future version of numpy, it will be understood as type, 1, / ‘1,type

1.软件环境

Windows10 教育版64位
Tensorflow-GPU 2.3.0
Keras 2.4.3
CUDA 10.1
Pyinstaller 4.10

2.问题描述

在使用 Tensorflow时，经常会打印一些无用的信息，主要包括如下：

2.1. CUDA 信息以及 CPU 指令集不支持 AVX2

Tensorflow在成功加载 CUDA后，会打印 CUDA相关的信息，真的是一片红~~：

2022-05-25 18:32:09.641749: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library cudart64_101.dll
2022-05-25 18:32:16.164069: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library nvcuda.dll
2022-05-25 18:32:16.186774: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1716] Found device 0 with properties:
pciBusID: 0000:73:00.0 name: Quadro RTX 6000 computeCapability: 7.5
coreClock: 1.77GHz coreCount: 72 deviceMemorySize: 24.00GiB deviceMemoryBandwidth: 625.94GiB/s
2022-05-25 18:32:16.187080: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library cudart64_101.dll
2022-05-25 18:32:16.190454: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library cublas64_10.dll
2022-05-25 18:32:16.193652: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library cufft64_10.dll
2022-05-25 18:32:16.194809: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library curand64_10.dll
2022-05-25 18:32:16.198553: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library cusolver64_10.dll
2022-05-25 18:32:16.200484: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library cusparse64_10.dll
2022-05-25 18:32:16.207734: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library cudnn64_7.dll
2022-05-25 18:32:16.207897: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1858] Adding visible gpu devices: 0
2022-05-25 18:32:17.342370: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN)to use the following CPU instructions in performance-critical operations:  AVX2
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.

2022-05-25 18:32:17.351683: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x28cb2077ca0 initialized for platform Host (this does not guarantee that XLA will be used). Devices:
2022-05-25 18:32:17.351949: I tensorflow/compiler/xla/service/service.cc:176]   StreamExecutor device (0): Host, Default Version
2022-05-25 18:32:17.352268: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1716] Found device 0 with properties:
pciBusID: 0000:73:00.0 name: Quadro RTX 6000 computeCapability: 7.5
coreClock: 1.77GHz coreCount: 72 deviceMemorySize: 24.00GiB deviceMemoryBandwidth: 625.94GiB/s
2022-05-25 18:32:17.352595: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library cudart64_101.dll
2022-05-25 18:32:17.352746: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library cublas64_10.dll
2022-05-25 18:32:17.352907: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library cufft64_10.dll
2022-05-25 18:32:17.353069: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library curand64_10.dll
2022-05-25 18:32:17.353242: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library cusolver64_10.dll
2022-05-25 18:32:17.353354: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library cusparse64_10.dll
2022-05-25 18:32:17.353448: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library cudnn64_7.dll
2022-05-25 18:32:17.353661: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1858] Adding visible gpu devices: 0
2022-05-25 18:32:17.881639: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1257] Device interconnect StreamExecutor with strength 1 edge matrix:
2022-05-25 18:32:17.881807: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1263]      0
2022-05-25 18:32:17.881871: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1276] 0:   N
2022-05-25 18:32:17.882136: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1402] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 21845 MB memory) -> physical GPU (device: 0, name: Quadro RTX 6000, pci bus id: 0000:73:00.0, compute capability: 7.5)
2022-05-25 18:32:17.885199: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x28ce667ab30 initialized for platform CUDA (this does not guarantee that XLA will be used). Devices:
2022-05-25 18:32:17.885412: I tensorflow/compiler/xla/service/service.cc:176]   StreamExecutor device (0): Quadro RTX 6000, Compute Capability 7.5

2.2. Tensorflow 1.0 和 Tensorflow 2.0 之间的 API 变更等 Warning

由于 Tensorflow 1.0和 Tensorflow 2.0之间的API变化较大，会打印许多 Warning信息：

WARNING:tensorflow:From C:\Users\Jayce\Anaconda3\envs\map_scale_gui\lib\site-packages\tensorflow\python\util\deprecation.py:574: calling map_fn_v2 (from tensorflow.python.ops.map_fn) with dtype is deprecated and will be removed in a future version.

Instructions for updating:
Use fn_output_signature instead
WARNING:tensorflow:From C:\Users\Jayce\Anaconda3\envs\map_scale_gui\lib\site-packages\tensorflow\python\keras\engine\training_v1.py:2070: Model.state_updates (from tensorflow.python.keras.engine.training) is deprecated and will be removed in a future version.

Instructions for updating:
This property should not be used in TensorFlow 2.0, as updates are applied automatically.

2.3. Pyinstaller 打包为exe后额外增加的 Warning

在使用 Pyinstaller等软件打包之后，程序还会额外打印一些新的警告：

WARNING:tensorflow:AutoGraph could not transform <bound method BatchNorm.call of <model.BatchNorm object at 0x0000021764BB9F60>> and will run it as-is.

Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, export AUTOGRAPH_VERBOSITY=10) and attach the full output.

Cause: Unable to locate the source code of <bound method BatchNorm.call of <model.BatchNorm object at 0x0000021764BB9F60>>. Note that functions defined in certain environments, like the interactive Python shell do not expose their source code. If that is the case, you should to define them in a .py source file. If you are certain the code is graph-compatible, wrap the call using @tf.autograph.do_not_convert. Original error: could not get source code
To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert
WARNING:tensorflow:AutoGraph could not transform <bound method ProposalLayer.call of <model.ProposalLayer object at 0x0000021767932208>> and will run it as-is.

Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, export AUTOGRAPH_VERBOSITY=10) and attach the full output.

Cause: Unable to locate the source code of <bound method ProposalLayer.call of <model.ProposalLayer object at 0x0000021767932208>>. Note that functions defined in certain environments, like the interactive Python shell do not expose their source code. If that is the case, you should to define them in a .py source file. If you are certain the code is graph-compatible, wrap the call using @tf.autograph.do_not_convert. Original error: could not get source code
To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert
WARNING:tensorflow:AutoGraph could not transform <bound method PyramidROIAlign.call of <model.PyramidROIAlign object at 0x00000217679A5860>> and will run it as-is.

Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, export AUTOGRAPH_VERBOSITY=10) and attach the full output.

Cause: Unable to locate the source code of <bound method PyramidROIAlign.call of <model.PyramidROIAlign object at 0x00000217679A5860>>. Note that functions defined in certain environments, like the interactive Python shell do not expose their source code. If that is the case, you should to define them in a .py source file. If you are certain the code is graph-compatible, wrap the call using @tf.autograph.do_not_convert. Original error: could not get source code
To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert
WARNING:tensorflow:AutoGraph could not transform <bound method DetectionLayer.call of <model.DetectionLayer object at 0x0000021767AB0320>> and will run it as-is.

Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, export AUTOGRAPH_VERBOSITY=10) and attach the full output.

Cause: Unable to locate the source code of <bound method DetectionLayer.call of <model.DetectionLayer object at 0x0000021767AB0320>>. Note that functions defined in certain environments, like the interactive Python shell do not expose their source code. If that is the case, you should to define them in a .py source file. If you are certain the code is graph-compatible, wrap the call using @tf.autograph.do_not_convert. Original error: could not get source code
To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert
WARNING:tensorflow:From tensorflow\python\util\deprecation.py:574: calling map_fn_v2 (from tensorflow.python.ops.map_fn) with dtype is deprecated and will be removed in a future version.

Instructions for updating:
Use fn_output_signature instead
WARNING:tensorflow:AutoGraph could not transform <function refine_detections_graph.<locals>.nms_keep_map at 0x0000021767AD99D8> and will run it as-is.

Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, export AUTOGRAPH_VERBOSITY=10) and attach the full output.

Cause: Unable to locate the source code of <function refine_detections_graph.<locals>.nms_keep_map at 0x0000021767AD99D8>. Note that functions defined in certain environments, like the interactive Python shell do not expose their source code. If that is the case, you should to define them in a .py source file. If you are certain the code is graph-compatible, wrap the call using @tf.autograph.do_not_convert. Original error: could not get source code
To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert

虽然上述信息不会影响我们程序的正常执行，但它有以下缺点：

• 如果你是多进程运行预测的话，上述信息会打印很多遍，这个时候你如果要捕获 CMD的输出，就会变得非常繁琐；
• 如果您的程序本身有需要打印的关键信息，上述输出很容易掩盖您需要查看的关键信息。
  
  [En]

  if your program itself has key information that needs to be printed, the above output can easily obscure the key information you need to view.*

• 无用输出过多可能影响程序性能
  
  [En]

  too much useless output may affect program performance*

对我们来说，其实上面的信息是 无用的。那么，有没有办法可以让我们将上述 Tensorflow中的无用信息屏蔽掉呢？

3.解决方法

3.1.取消打印 CUDA 信息以及 CPU 指令集不支持 AVX2

TensorFlow中的 Log共有 DEBUG、 INFO、 WARNING、 ERROR、 FATAL 5种级别，且通过环境变量 TF_CPP_MIN_LOG_LEVEL对日志级别进行设置， TF_CPP_MIN_LOG_LEVEL不同值的含义如下：

LevelLevel for HumansLevel Description0DEBUGall messages are logged (Default)1INFOINFO messages are not printed2WARNINGINFO and WARNING messages are not printed3ERRORINFO, WARNING, and ERROR messages are not printed

因此我们可以通过设置 TF_CPP_MIN_LOG_LEVEL的值，屏蔽该级别以及更低级别的日志，如：设置为 3，表示屏蔽 INFO, WARNING, ERROR级别的日志。

3.1.1.具体操作

直接在你的主函数（即你运行的 .py文件）开头添加如下代码：

import os
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'

记住一定是 代码开头，如果你把上面代码放在 import tensorflow as tf后面，这个时候 tensorflow都 import完了才会运行我们的语句，自然设置就不会有效果了。

3.1.2.效果预览

这个时候，我们已经成功取消打印 CUDA信息以及 CPU指令集不支持 AVX2等无用信息，之前界面一大堆红色的警告信息已经消失不见，只剩下 API变更的警告：

; 3.2.取消打印 Tensorflow 2.0 API变更以及 Pyinstaller 打包为 exe 后额外增加的 Warning

Tensorflow为了兼容 1.x和 2.x 的 APIs，专门提供了 tf.compat模块：

3.2.1.具体操作

直接在 3.1添加的代码后面继续添加如下代码：

import tensorflow as tf
tf.compat.v1.logging.set_verbosity(tf.compat.v1.logging.ERROR)

3.2.2.效果预览

这个时候， 3.1剩下的 API变更的警告也没有了，控制台只打印我们自己定义的提示信息：

; 4.总结

如果想完全消除 Tensorflow本身打印的警告，需要结合 3.1和 3.2，即直接在你的主函数（即你运行的 .py文件）开头添加代码：

import os
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'
import tensorflow as tf
tf.compat.v1.logging.set_verbosity(tf.compat.v1.logging.ERROR)

如果只想屏蔽一部分，如只想取消打印 CUDA信息以及 CPU指令集不支持 AVX2，则按 3.1操作即可；而如果只想取消打印 Tensorflow 2.0的 API变更以及 Pyinstaller打包为 exe后额外增加的 Warning，则按 3.2操作即可。但一定要注意：是在你即将运行的 .py文件的 开头添加相应代码。

​

都看到这里了，还不赶紧 点赞， 评论， 收藏走一波？

Original: https://blog.csdn.net/qq_15969343/article/details/124971292
Author: 任博今天就要毕业
Title: 取消打印Tensorflow中的无用信息，如tensorflow:AutoGraph could not transform ＜*＞ and will run it as-is、加载CUDA信息等