可以分别使用 GPU 和 CPU 运行同一个模型来选择合适的运算设备。GPU 和 CPU 的选择与模型类型也有关系，例如如果是卷积神经网络则 gtx1050 可能胜过 i7，但如果是强化学习则不一定。
可以简单地认为，是的。不需要两个版本都装，TensorFlow 2.1 默认安装（pip install tensorflow）就已经同时支持 CPU 和 GPU，可以使用
cpus = tf.config.list_physical_devices (device_type=‘CPU’)
tf.config.set_visible_devices (devices=cpus)
来限定只使用 CPU 进行运算。
1、tf.config.set_visible_devices (devices=cpus) 设定似乎有问题，搜索了 Tf2 官方，也没看明白这个用法。
2、一种方法，类似于当前程序的全局设定，放在 import tensorflow 之前
import os
os.environ [“CUDA_DEVICE_ORDER”] = “PCI_BUS_ID”
os.environ [“CUDA_VISIBLE_DEVICES”] = “-1”
3，另一种方法，是程序临时指定
with tf.device (’/cpu:0’):
A = tf.constant ([[1, 2], [3, 4]])
B = tf.constant ([[5, 6], [7, 8]])
C = tf.matmul (A, B)
print ©
with tf.device (’/gpu:0’):
A = tf.constant ([[1, 2], [3, 4]])
B = tf.constant ([[5, 6], [7, 8]])
C = tf.matmul (A, B)
print ©
可以参考 tf.config.set_visible_devices | TensorFlow v2.14.0 ，相关内容在 https://tf.wiki/zh/basic/tools.html#tf-config-gpu 也有介绍。一个简单的示例程序如下：
import tensorflow as tf
tf.debugging.set_log_device_placement (True) # 设置输出运算所在的设备
cpus = tf.config.list_physical_devices ('CPU') # 获取当前设备的 CPU 列表
tf.config.set_visible_devices (cpus) # 设置 TensorFlow 的可见设备范围为 cpu
A = tf.constant ([[1, 2], [3, 4]])
B = tf.constant ([[5, 6], [7, 8]])
C = tf.matmul (A, B)
print (C)
输出
2020-04-21 11:37:29.007897: I tensorflow/core/common_runtime/eager/execute.cc:573] Executing op MatMul in device /job:localhost/replica:0/task:0/device:CPU:0
tf.Tensor (
[[19 22]
[43 50]], shape=(2, 2), dtype=int32)
当然最保险的办法是新开一个 conda 虚拟环境并且
pip install tensorflow-cpu
，安装仅支持 CPU 的 TensorFlow 版本。
看起来可能是网络问题。如果使用 conda 安装的话建议设置镜像。
When I use pip install tensorflow
I got the feedback:
ERROR: Could not find a version that satisfies the requirement tensorflow (from versions: none)
ERROR: No matching distribution found for tensorflow
But I find the command worked:
python -m pip install --upgrade https://storage.googleapis.com/tensorflow/mac/cpu/tensorflow-0.12.0-py3-none-any.whl
If you meet such a question, you can try it.
好吧我找到了原因了
安装一直出错是因为我之前下了 Python3.7.4-32bit 版本，换成了 64bit 就没有问题了
关于 IDE 设置 的，我的路径是 /opt/anaconda3/envs/tf2/bin/python ，不知道会不会有问题……。我系统是 macOS 10.15.3 (19D76)
没有问题，我与你的一致
你好，按照这种方式在 Anaconda 环境中使用 TF2.1 会遇到在使用 Tensorboard 的 Profile 时报错的问题，报错如下：
2020-05-08 11:09:24.374761: W tensorflow/stream_executor/platform/default/dso_loader.cc:55] Could not load dynamic library 'libcupti.so.10.1'; dlerror: libcupti.so.10.1: cannot open shared object file: No such file or directory; LD_LIBRARY_PATH: /opt/gridview//pbs/dispatcher/lib::/usr/local/lib64:/usr/local/lib
2020-05-08 11:09:24.374801: E tensorflow/core/profiler/internal/gpu/cupti_tracer.cc:1307] function cupti_interface_->Subscribe ( &subscriber_, (CUpti_CallbackFunc) ApiCallback, this) failed with error CUPTI could not be loaded or symbol could not be found.
2020-05-08 11:09:24.374816: E tensorflow/core/profiler/internal/gpu/cupti_tracer.cc:1346] function cupti_interface_->ActivityRegisterCallbacks ( AllocCuptiActivityBuffer, FreeCuptiActivityBuffer) failed with error CUPTI could not be loaded or symbol could not be found.
另附 nvidia-smi 和 conda list 的结果：
Fri May 8 11:25:59 2020
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 430.50 Driver Version: 430.50 CUDA Version: 10.1 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 Tesla V100-PCIE... Off | 00000000:89:00.0 Off | 0 |
| N/A 40C P0 37W / 250W | 1354MiB / 32510MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
| 1 Tesla V100-PCIE... Off | 00000000:8A:00.0 Off | 0 |
| N/A 39C P0 38W / 250W | 320MiB / 32510MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
| 2 Tesla V100-PCIE... Off | 00000000:8B:00.0 Off | 0 |
| N/A 39C P0 36W / 250W | 320MiB / 32510MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
| 3 Tesla V100-PCIE... Off | 00000000:8C:00.0 Off | 0 |
| N/A 43C P0 49W / 250W | 18828MiB / 32510MiB | 65% Default |
+-------------------------------+----------------------+----------------------+
| 4 Tesla V100-PCIE... Off | 00000000:DA:00.0 Off | 0 |
| N/A 40C P0 46W / 250W | 18828MiB / 32510MiB | 73% Default |
+-------------------------------+----------------------+----------------------+
| 5 Tesla V100-PCIE... Off | 00000000:DB:00.0 Off | 0 |
| N/A 38C P0 39W / 250W | 320MiB / 32510MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
| 6 Tesla V100-PCIE... Off | 00000000:DC:00.0 Off | 0 |
| N/A 64C P0 155W / 250W | 5789MiB / 32510MiB | 86% Default |
+-------------------------------+----------------------+----------------------+
| 7 Tesla V100-PCIE... Off | 00000000:DD:00.0 Off | 0 |
| N/A 40C P0 40W / 250W | 320MiB / 32510MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
| 0 2234 G /usr/bin/X 22MiB |
| 0 83392 C /usr/lhy/anaconda3/envs/tf2/bin/python 363MiB |
| 0 135660 C /usr/lhy/anaconda3/envs/tf2/bin/python 955MiB |
| 1 135660 C /usr/lhy/anaconda3/envs/tf2/bin/python 307MiB |
| 2 135660 C /usr/lhy/anaconda3/envs/tf2/bin/python 307MiB |
| 3 135660 C /usr/lhy/anaconda3/envs/tf2/bin/python 307MiB |
| 3 139517 C python 4627MiB |
| 3 140633 C python 4627MiB |
| 3 142274 C python 4627MiB |
| 3 143266 C python 4627MiB |
| 4 133393 C python 4627MiB |
| 4 133906 C python 4627MiB |
| 4 134679 C python 4627MiB |
| 4 135379 C python 4627MiB |
| 4 135660 C /usr/lhy/anaconda3/envs/tf2/bin/python 307MiB |
| 5 135660 C /usr/lhy/anaconda3/envs/tf2/bin/python 307MiB |
| 6 15156 C python 5469MiB |
| 6 135660 C /usr/lhy/anaconda3/envs/tf2/bin/python 307MiB |
| 7 135660 C /usr/lhy/anaconda3/envs/tf2/bin/python 307MiB |
+-----------------------------------------------------------------------------+
# packages in environment at /usr/lhy/anaconda3/envs/tf2:
#
# Name Version Build Channel
_libgcc_mutex 0.1 main defaults
absl-py 0.9.0 pypi_0 pypi
astor 0.8.1 pypi_0 pypi
attrs 19.3.0 pypi_0 pypi
backcall 0.1.0 py37_0 defaults
bleach 3.1.4 pypi_0 pypi
ca-certificates 2020.1.1 0 defaults
cachetools 4.1.0 pypi_0 pypi
certifi 2020.4.5.1 py37_0 defaults
chardet 3.0.4 pypi_0 pypi
cloudpickle 1.3.0 pypi_0 pypi
cudatoolkit 10.1.243 h6bb024c_0 defaults
cudnn 7.6.5 cuda10.1_0 defaults
cycler 0.10.0 pypi_0 pypi
decorator 4.4.2 py_0 defaults
defusedxml 0.6.0 pypi_0 pypi
entrypoints 0.3 pypi_0 pypi
gast 0.2.2 pypi_0 pypi
gin-config 0.1.3 pypi_0 pypi
google-auth 1.14.0 pypi_0 pypi
google-auth-oauthlib 0.4.1 pypi_0 pypi
google-pasta 0.2.0 pypi_0 pypi
grpcio 1.28.1 pypi_0 pypi
gym 0.10.11 pypi_0 pypi
h5py 2.10.0 pypi_0 pypi
idna 2.9 pypi_0 pypi
importlib-metadata 1.6.0 pypi_0 pypi
ipykernel 5.2.1 pypi_0 pypi
ipython 7.13.0 py37h5ca1d4c_0 defaults
ipython-genutils 0.2.0 pypi_0 pypi
ipython_genutils 0.2.0 py37_0 defaults
jedi 0.17.0 pypi_0 pypi
jinja2 2.11.2 pypi_0 pypi
joblib 0.14.1 pypi_0 pypi
json5 0.9.4 pypi_0 pypi
jsonschema 3.2.0 pypi_0 pypi
jupyter-client 6.1.3 pypi_0 pypi
jupyter_client 6.1.2 py_0 defaults
jupyter_core 4.6.3 py37_0 defaults
jupyterlab 2.1.0 pypi_0 pypi
jupyterlab-server 1.1.1 pypi_0 pypi
keras-applications 1.0.8 pypi_0 pypi
keras-preprocessing 1.1.0 pypi_0 pypi
kiwisolver 1.2.0 pypi_0 pypi
ld_impl_linux-64 2.33.1 h53a641e_7 defaults
libedit 3.1.20181209 hc058e9b_0 defaults
libffi 3.2.1 hd88cf55_4 defaults
libgcc-ng 9.1.0 hdf63c60_0 defaults
libsodium 1.0.16 h1bed415_0 defaults
libstdcxx-ng 9.1.0 hdf63c60_0 defaults
lightgbm 2.3.1 pypi_0 pypi
markdown 3.2.1 pypi_0 pypi
markupsafe 1.1.1 pypi_0 pypi
matplotlib 3.2.1 pypi_0 pypi
minepy 1.2.4 pypi_0 pypi
mistune 0.8.4 pypi_0 pypi
nbconvert 5.6.1 pypi_0 pypi
nbformat 5.0.6 pypi_0 pypi
ncurses 6.2 he6710b0_0 defaults
notebook 6.0.3 pypi_0 pypi
numpy 1.18.3 pypi_0 pypi
oauthlib 3.1.0 pypi_0 pypi
openssl 1.1.1g h7b6447c_0 defaults
opt-einsum 3.2.1 pypi_0 pypi
pandas 1.0.3 pypi_0 pypi
pandocfilters 1.4.2 pypi_0 pypi
parso 0.7.0 pypi_0 pypi
pexpect 4.8.0 py37_0 defaults
pickleshare 0.7.5 pypi_0 pypi
pip 20.0.2 py37_1 defaults
prometheus-client 0.7.1 pypi_0 pypi
prompt-toolkit 3.0.5 pypi_0 pypi
prompt_toolkit 3.0.4 0 defaults
protobuf 3.11.3 pypi_0 pypi
ptyprocess 0.6.0 pypi_0 pypi
pyaml 20.4.0 pypi_0 pypi
pyasn1 0.4.8 pypi_0 pypi
pyasn1-modules 0.2.8 pypi_0 pypi
pydot 1.4.1 pypi_0 pypi
pyglet 1.5.4 pypi_0 pypi
pygments 2.6.1 py_0 defaults
pyparsing 2.4.7 pypi_0 pypi
pyrsistent 0.16.0 pypi_0 pypi
python 3.7.7 hcf32534_0_cpython defaults
python-dateutil 2.8.1 py_0 defaults
python-graphviz 0.14 pypi_0 pypi
pytz 2019.3 pypi_0 pypi
pyyaml 5.3.1 pypi_0 pypi
pyzmq 19.0.0 pypi_0 pypi
readline 8.0 h7b6447c_0 defaults
requests 2.23.0 pypi_0 pypi
requests-oauthlib 1.3.0 pypi_0 pypi
rsa 4.0 pypi_0 pypi
scikit-learn 0.22.2.post1 pypi_0 pypi
scipy 1.4.1 pypi_0 pypi
send2trash 1.5.0 pypi_0 pypi
setuptools 46.1.3 py37_0 defaults
six 1.14.0 py37_0 defaults
SQLite 3.31.1 h7b6447c_0 defaults
tensorboard 2.1.1 pypi_0 pypi
tensorflow 2.1.0 pypi_0 pypi
tensorflow-addons 0.9.1 pypi_0 pypi
tensorflow-estimator 2.1.0 pypi_0 pypi
tensorflow-probability 0.9.0 pypi_0 pypi
termcolor 1.1.0 pypi_0 pypi
terminado 0.8.3 pypi_0 pypi
testpath 0.4.4 pypi_0 pypi
tf-agents 0.4.0 pypi_0 pypi
tk 8.6.8 hbc83047_0 defaults
tornado 6.0.4 py37h7b6447c_1 defaults
tqdm 4.45.0 pypi_0 pypi
traitlets 4.3.3 py37_0 defaults
typeguard 2.7.1 pypi_0 pypi
urllib3 1.25.9 pypi_0 pypi
wcwidth 0.1.9 py_0 defaults
webencodings 0.5.1 pypi_0 pypi
werkzeug 1.0.1 pypi_0 pypi
wheel 0.34.2 py37_0 defaults
wrapt 1.12.1 pypi_0 pypi
xgboost 0.80 pypi_0 pypi
xz 5.2.5 h7b6447c_0 defaults
zeromq 4.3.1 he6710b0_3 defaults
zipp 3.1.0 pypi_0 pypi
zlib 1.2.11 h7b6447c_3 defaults
ERROR: Could not find a version that satisfies the requirement tensorflow (from versions: none)
ERROR: No matching distribution found for tensorflow
按你说的装不了 tensorflow
1 reply
您好！我安装的 tf 2.0，每次启动时，cuda 的加载信息如下，有的信息重复显示了两次，但是计算上没有问题，请问您知道是什么原因吗？
2020-05-22 09:09:35.094679: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcuda.so.1
2020-05-22 09:09:35.146396: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1618] Found device 0 with properties:
name: TITAN X (Pascal) major: 6 minor: 1 memoryClockRate (GHz): 1.531
pciBusID: 0000:02:00.0
2020-05-22 09:09:35.147109: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1618] Found device 1 with properties:
name: Quadro P2000 major: 6 minor: 1 memoryClockRate (GHz): 1.4805
pciBusID: 0000:01:00.0
2020-05-22 09:09:35.147377: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcudart.so.10.0
2020-05-22 09:09:35.149235: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcublas.so.10.0
2020-05-22 09:09:35.150833: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcufft.so.10.0
2020-05-22 09:09:35.151172: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcurand.so.10.0
2020-05-22 09:09:35.153255: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcusolver.so.10.0
2020-05-22 09:09:35.154837: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcusparse.so.10.0
2020-05-22 09:09:35.159338: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcudnn.so.7
2020-05-22 09:09:35.162403: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1746] Adding visible gpu devices: 0, 1
2020-05-22 09:09:35.198717: I tensorflow/core/platform/profile_utils/cpu_utils.cc:94] CPU Frequency: 2194750000 Hz
2020-05-22 09:09:35.201895: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x56304c7438d0 executing computations on platform Host. Devices:
2020-05-22 09:09:35.201939: I tensorflow/compiler/xla/service/service.cc:175] StreamExecutor device (0): Host, Default Version
2020-05-22 09:09:35.545476: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x56304ca88560 executing computations on platform CUDA. Devices:
2020-05-22 09:09:35.545540: I tensorflow/compiler/xla/service/service.cc:175] StreamExecutor device (0): TITAN X (Pascal), Compute Capability 6.1
2020-05-22 09:09:35.545550: I tensorflow/compiler/xla/service/service.cc:175] StreamExecutor device (1): Quadro P2000, Compute Capability 6.1
2020-05-22 09:09:35.547244: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1618] Found device 0 with properties:
name: TITAN X (Pascal) major: 6 minor: 1 memoryClockRate (GHz): 1.531
pciBusID: 0000:02:00.0
2020-05-22 09:09:35.548178: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1618] Found device 1 with properties:
name: Quadro P2000 major: 6 minor: 1 memoryClockRate (GHz): 1.4805
pciBusID: 0000:01:00.0
2020-05-22 09:09:35.548249: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcudart.so.10.0
2020-05-22 09:09:35.548280: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcublas.so.10.0
2020-05-22 09:09:35.548307: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcufft.so.10.0
2020-05-22 09:09:35.548333: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcurand.so.10.0
2020-05-22 09:09:35.548359: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcusolver.so.10.0
2020-05-22 09:09:35.548384: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcusparse.so.10.0
2020-05-22 09:09:35.548411: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcudnn.so.7
2020-05-22 09:09:35.552360: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1746] Adding visible gpu devices: 0, 1
2020-05-22 09:09:35.552423: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcudart.so.10.0
2020-05-22 09:09:35.555863: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1159] Device interconnect StreamExecutor with strength 1 edge matrix:
2020-05-22 09:09:35.555894: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1165] 0 1
2020-05-22 09:09:35.555908: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1178] 0: N N
2020-05-22 09:09:35.555919: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1178] 1: N N
2020-05-22 09:09:35.559429: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1304] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 11435 MB memory) -> physical GPU (device: 0, name: TITAN X (Pascal), pci bus id: 0000:02:00.0, compute capability: 6.1)
2020-05-22 09:09:35.560428: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1304] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:1 with 4518 MB memory) -> physical GPU (device: 1, name: Quadro P2000, pci bus id: 0000:01:00.0, compute capability: 6.1)
<tf.Tensor: id=0, shape=(), dtype=int32, numpy=1>
同问，安装完 tensorflow2.2，也是一样的，但是输出过程正常
@MingCheung @hifeng2017 TensorFlow 在运行过程中会在终端打一些 log，这个是正常的，具体使用上只要不报错就行（有 warning 的话大多是使用了过时的 api，可以根据提示做一些调整）
如果不喜欢这些 log，可以参考 https://stackoverflow.com/questions/35911252/disable-tensorflow-debugging-information 把 log 关掉。
我没有用过 spider，所以只能建议你在网络上搜索一下错误信息。本手册建议使用 PyCharm 作为 IDE，参考 https://tf.wiki/zh_hans/basic/installation.html#ide
我在 import tensorflow as tf 时出现下面的错误是什么问题，之前装也没有问题。
2020-06-28 21:14:21.568630: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudart64_101.dll
Traceback (most recent call last):
File “”, line 1, in
File “F:\Python\InstallPath\lib\site-packages\tensorflow_init_.py”, line 41, in
from tensorflow.python.tools import module_util as module_util
File "F:\Python\InstallPath\lib\site-packages\tensorflow\python_init.py", line 84, in
from tensorflow.python import keras
File “F:\Python\InstallPath\lib\site-packages\tensorflow\python\keras_init_.py”, line 27, in
from tensorflow.python.keras import models
File “F:\Python\InstallPath\lib\site-packages\tensorflow\python\keras\models.py”, line 27, in
from tensorflow.python.keras.engine import sequential
File “F:\Python\InstallPath\lib\site-packages\tensorflow\python\keras\engine\sequential.py”, line 27, in
from tensorflow.python.keras.engine import training
File “F:\Python\InstallPath\lib\site-packages\tensorflow\python\keras\engine\training.py”, line 37, in
from tensorflow.python.keras.engine import data_adapter
File “F:\Python\InstallPath\lib\site-packages\tensorflow\python\keras\engine\data_adapter.py”, line 55, in
from scipy import sparse as scipy_sparse # pylint: disable=g-import-not-at-top
File “F:\Python\InstallPath\lib\site-packages\scipy_init_.py”, line 106, in
from . import distributor_init
File “F:\Python\InstallPath\lib\site-packages\scipy_distributor_init.py”, line 26, in
WinDLL (os.path.abspath (filename))
File "F:\Python\InstallPath\lib\ctypes_init.py", line 364, in init
self._handle = _dlopen (self._name, mode)
OSError: [WinError 126] 找不到指定的模块。
劳烦说明一下安装的 TensorFlow 版本，以及请参考 https://tf.wiki/zh/basic/installation.html#id12 中的 “导入 TensorFlow 时部分可能出现的错误信息及解决方案”。考虑安装一下 Microsoft Visual C++ Redistributable for Visual Studio 2015, 2017 and 2019 看看。
1 reply
刚开始安装的是 GPU 版本 2.2.0，cuda10.1，cudnn7.6.5。后来又降级到 2.1.0 版本，还是这个问题。
运行的代码是 print (tf.reduce_sum (tf.random.normal ([10, 10])))，运行时间都要 10 分钟，我的电脑是 cpui5，gpu 是 gtx950M，我之前装过 tensorflow2.2，运行情况不是这样的。
完整的终端输出内容是：
2020-07-27 17:28:44.898662: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcuda.so.1
2020-07-27 17:28:45.943645: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2020-07-27 17:28:45.945260: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1561] Found device 0 with properties:
pciBusID: 0000:01:00.0 name: GeForce GTX 950M computeCapability: 5.0
coreClock: 1.124GHz coreCount: 5 deviceMemorySize: 3.95GiB deviceMemoryBandwidth: 29.83GiB/s
2020-07-27 17:28:45.993423: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcudart.so.10.1
2020-07-27 17:28:46.454724: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcublas.so.10
2020-07-27 17:28:46.789430: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcufft.so.10
2020-07-27 17:28:46.891430: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcurand.so.10
2020-07-27 17:28:47.415397: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcusolver.so.10
2020-07-27 17:28:47.655817: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcusparse.so.10
2020-07-27 17:28:48.329574: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcudnn.so.7
2020-07-27 17:28:48.329881: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2020-07-27 17:28:48.330447: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2020-07-27 17:28:48.330787: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1703] Adding visible gpu devices: 0
2020-07-27 17:28:48.356413: I tensorflow/core/platform/cpu_feature_guard.cc:143] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA
2020-07-27 17:28:48.577678: I tensorflow/core/platform/profile_utils/cpu_utils.cc:102] CPU Frequency: 2299965000 Hz
2020-07-27 17:28:48.577952: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x4dd02d0 initialized for platform Host (this does not guarantee that XLA will be used). Devices:
2020-07-27 17:28:48.577971: I tensorflow/compiler/xla/service/service.cc:176] StreamExecutor device (0): Host, Default Version
2020-07-27 17:28:49.139140: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2020-07-27 17:28:49.139682: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x4dd2e20 initialized for platform CUDA (this does not guarantee that XLA will be used). Devices:
2020-07-27 17:28:49.139747: I tensorflow/compiler/xla/service/service.cc:176] StreamExecutor device (0): GeForce GTX 950M, Compute Capability 5.0
2020-07-27 17:28:49.173354: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2020-07-27 17:28:49.176437: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1561] Found device 0 with properties:
pciBusID: 0000:01:00.0 name: GeForce GTX 950M computeCapability: 5.0
coreClock: 1.124GHz coreCount: 5 deviceMemorySize: 3.95GiB deviceMemoryBandwidth: 29.83GiB/s
2020-07-27 17:28:49.194079: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcudart.so.10.1
2020-07-27 17:28:49.194111: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcublas.so.10
2020-07-27 17:28:49.194127: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcufft.so.10
2020-07-27 17:28:49.194144: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcurand.so.10
2020-07-27 17:28:49.194160: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcusolver.so.10
2020-07-27 17:28:49.194175: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcusparse.so.10
2020-07-27 17:28:49.194191: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcudnn.so.7
2020-07-27 17:28:49.194313: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2020-07-27 17:28:49.194822: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2020-07-27 17:28:49.195171: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1703] Adding visible gpu devices: 0
2020-07-27 17:28:49.220996: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcudart.so.10.1
2020-07-27 17:28:49.236580: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1102] Device interconnect StreamExecutor with strength 1 edge matrix:
2020-07-27 17:28:49.236624: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1108] 0
2020-07-27 17:28:49.236634: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1121] 0: N
2020-07-27 17:28:49.245993: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2020-07-27 17:28:49.246740: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2020-07-27 17:28:49.247191: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1247] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 3708 MB memory) -> physical GPU (device: 0, name: GeForce GTX 950M, pci bus id: 0000:01:00.0, compute capability: 5.0)
输出内容有点长，麻烦您帮我看一下
踩过坑，留给后面朋友的建议。
macOS catelina 10.15.4.
Anaconda 环境的理解 初学 Python 者自学 Anaconda 的正确姿势是什么？ - 知乎
使用 conda 更新源修复 tensofflow 安装过程中 wrap 出现的问题 ERROR: Cannot uninstall 'wrapt'. during upgrade · Issue #30191 · tensorflow/tensorflow · GitHub → conda update wrapt
pip install tensorflow 需要针对虚拟环境进行设置，而不是宿主环境
在 PyCharm 中运行代码后，出现以下代码，但是没有报错，输出结果正常。请问是什么原因？
2020-08-08 18:02:52.694491: W tensorflow/stream_executor/platform/default/dso_loader.cc:59] Could not load dynamic library ‘cudart64_101.dll’; dlerror: cudart64_101.dll not found
2020-08-08 18:02:52.694817: I tensorflow/stream_executor/cuda/cudart_stub.cc:29] Ignore above cudart dlerror if you do not have a GPU set up on your machine.
1 reply2020-08-08 18:02:57.836214: W tensorflow/stream_executor/platform/default/dso_loader.cc:59] Could not load dynamic library ‘nvcuda.dll’; dlerror: nvcuda.dll not found
2020-08-08 18:02:57.836522: W tensorflow/stream_executor/cuda/cuda_driver.cc:312] failed call to cuInit: UNKNOWN ERROR (303)
2020-08-08 18:02:57.849447: I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:169] retrieving CUDA diagnostic information for host:
2020-08-08 18:02:57.849992: I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:176] hostname:
2020-08-08 18:02:57.850409: 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.
2020-08-08 18:02:57.862757: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x11aac2bc010 initialized for platform Host (this does not guarantee that XLA will be used). Devices:
2020-08-08 18:02:57.863116: I tensorflow/compiler/xla/service/service.cc:176] StreamExecutor device (0): Host, Default Version
运行途中可能会输出一些 TensorFlow 的提示信息，属于正常现象。
如果你指的是 cudart64_101.dll not found，说明 CUDA 未安装或未正确安装，可以参考手册的 “GPU 版本 TensorFlow 安装指南” https://tf.wiki/zh_hans/basic/installation.html#gputensorflow
import tensorflow as tf
A = tf.constant ([[1, 2], [3, 4]])
B = tf.constant ([[5, 6], [7, 8]])
C = tf.matmul (A, B)
print ©
为什么我的运行结果为：
Tensor (“MatMul_9:0”, shape=(2, 2), dtype=int32)
请问在使用 gpu 运行时出现这样的错误是什么原因（pycharm 的提示信息显示，程序能识别到 gpu，能够显示 gpu 的型号和算力）：tensorflow.python.framework.errors_impl.InvalidArgumentError: assertion failed: [0] [Op:Assert] name: EagerVariableNameReuse
这种情况出现在指定了 gpu 或者使用默认状态，如果切换成纯 cpu 处理则可以顺利执行代码。问题报错位置为 tf.keras.models.Sequential
求解答，万分感谢
import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import datasets, layers, optimizers, Sequential, metrics
import os
os.environ ['TF_CPP_MIN_LOG_LEVEL'] = "2"
os.environ ['CUDA_VISIBLE_DEVICES'] = '0'
def preprocess (x, y):
x = tf.cast (x, dtype=tf.float32)/255.0
y = tf.cast (y, dtype=tf.int32)
return x, y
(x, y), (x_test, y_test) = tf.keras.datasets.fashion_mnist.load_data ()
print (x.shape, y.shape, x_test.shape, y_test.shape)
batchsize = 128
db = tf.data.Dataset.from_tensor_slices ((x,y))
db = db.map (preprocess).shuffle (10000).batch (batchsize)
db_test = tf.data.Dataset.from_tensor_slices ((x_test,y_test))
db_test = db_test.map (preprocess).batch (batchsize)
db_iter = iter (db)
sample = next (db_iter)
print (sample [0].shape, sample [1].shape)
model = keras.Sequential ([
layers.Dense (256, activation=tf.nn.relu),
layers.Dense (128, activation=tf.nn.relu),
layers.Dense (64, activation=tf.nn.relu),
layers.Dense (32, activation=tf.nn.relu),
layers.Dense (10)
])
model.build (input_shape=[None, 28*28])
model.summary ()
optimizer = optimizers.Adam (lr=1e-3)
def main ():
for epoch in range (30):
for step, (x, y) in enumerate (db):
x = tf.reshape (x, [-1, 28*28])
with tf.GradientTape () as tape:
y_onehot = tf.one_hot (y,depth=10)
logits = model (x)
loss1 = tf.reduce_mean (tf.losses.MSE (y_onehot, logits))
loss2 = tf.reduce_mean (tf.losses.categorical_crossentropy (y_onehot, logits, from_logits=True))
grads = tape.gradient (loss1, model.trainable_variables)
optimizer.apply_gradients (zip (grads,model.trainable_variables))
if step % 100 == 0:
print (epoch, step, "loss:", float (loss1), float (loss2))
total_correct = 0
total_num = 0
for x,y in db_test:
x = tf.reshape (x, [-1, 28 * 28])
logits = model (x)
prob = tf.nn.softmax (logits, axis=1)
pred = tf.argmax (prob, axis=1)
pred = tf.cast (pred, dtype=tf.int32)
correct = tf.equal (pred, y)
correct = tf.reduce_sum (tf.cast (correct, dtype=tf.int32))
total_correct += int (correct)
total_num +=x.shape [0]
acc = total_correct / total_num
print (epoch, "acc:", acc)
if __name__ == '__main__':
main ()
Traceback (most recent call last):
File "E:/Python_code/test/FashionMnist_layer", line 30, in <module>
model = keras.Sequential ([
File "D:\Program\Python\lib\site-packages\tensorflow\python\training\tracking\base.py", line 457, in _method_wrapper
result = method (self, *args, **kwargs)
File "D:\Program\Python\lib\site-packages\tensorflow\python\keras\engine\sequential.py", line 116, in __init__
super (functional.Functional, self).__init__( # pylint: disable=bad-super-call
File "D:\Program\Python\lib\site-packages\tensorflow\python\training\tracking\base.py", line 457, in _method_wrapper
result = method (self, *args, **kwargs)
File "D:\Program\Python\lib\site-packages\tensorflow\python\keras\engine\training.py", line 308, in __init__
self._init_batch_counters ()
File "D:\Program\Python\lib\site-packages\tensorflow\python\training\tracking\base.py", line 457, in _method_wrapper
result = method (self, *args, **kwargs)
File "D:\Program\Python\lib\site-packages\tensorflow\python\keras\engine\training.py", line 317, in _init_batch_counters
self._train_counter = variables.Variable (0, dtype='int64', aggregation=agg)
File "D:\Program\Python\lib\site-packages\tensorflow\python\ops\variables.py", line 262, in __call__
return cls._variable_v2_call (*args, **kwargs)
File "D:\Program\Python\lib\site-packages\tensorflow\python\ops\variables.py", line 244, in _variable_v2_call
return previous_getter (
File "D:\Program\Python\lib\site-packages\tensorflow\python\ops\variables.py", line 237, in <lambda>
previous_getter = lambda **kws: default_variable_creator_v2 (None, **kws)
File "D:\Program\Python\lib\site-packages\tensorflow\python\ops\variable_scope.py", line 2633, in default_variable_creator_v2
return resource_variable_ops.ResourceVariable (
File "D:\Program\Python\lib\site-packages\tensorflow\python\ops\variables.py", line 264, in __call__
return super (VariableMetaclass, cls).__call__(*args, **kwargs)
File "D:\Program\Python\lib\site-packages\tensorflow\python\ops\resource_variable_ops.py", line 1507, in __init__
self._init_from_args (
File "D:\Program\Python\lib\site-packages\tensorflow\python\ops\resource_variable_ops.py", line 1661, in _init_from_args
handle = eager_safe_variable_handle (
File "D:\Program\Python\lib\site-packages\tensorflow\python\ops\resource_variable_ops.py", line 242, in eager_safe_variable_handle
return _variable_handle_from_shape_and_dtype (
File "D:\Program\Python\lib\site-packages\tensorflow\python\ops\resource_variable_ops.py", line 174, in _variable_handle_from_shape_and_dtype
gen_logging_ops._assert ( # pylint: disable=protected-access
File "D:\Program\Python\lib\site-packages\tensorflow\python\ops\gen_logging_ops.py", line 49, in _assert
_ops.raise_from_not_ok_status (e, name)
File "D:\Program\Python\lib\site-packages\tensorflow\python\framework\ops.py", line 6843, in raise_from_not_ok_status
six.raise_from (core._status_to_exception (e.code, message), None)
File "<string>", line 3, in raise_from
tensorflow.python.framework.errors_impl.InvalidArgumentError: assertion failed: [0] [Op:Assert] name: EagerVariableNameReuse这边在 Colab 的 GPU 上跑了一下没有问题，建议参考 InvalidArgumentError: assertion failed: [0] [Op:Assert] name: EagerVariableNameReuse · Issue #38518 · tensorflow/tensorflow · GitHub ，以及检查你的 GPU 是不是被其他程序占用了。
(tf2) C:\Users\Xavie>python
Python 3.7.9 (default, Aug 31 2020, 17:10:11) [MSC v.1916 64 bit (AMD64)] :: Anaconda, Inc. on win32
Type “help”, “copyright”, “credits” or “license” for more information.
import tensorflow as tf
2020-09-11 20:33:26.555854: W tensorflow/stream_executor/platform/default/dso_loader.cc:59] Could not load dynamic library ‘cudart64_101.dll’; dlerror: cudart64_101.dll not found
2020-09-11 20:33:26.560825: I tensorflow/stream_executor/cuda/cudart_stub.cc:29] Ignore above cudart dlerror if you do not have a GPU set up on your machine.
這樣了，請問如何解決的
1 reply
這個信息表明你的電腦沒有安裝支持CUDA的GPU或者GPU配置不正確，會使用CPU進行運算。如果你的電腦確實沒有NVIDIA的顯示卡或者不打算用GPU，可忽略此提示信息。
我在使用tensorflow时总是有这个warning，请问是什么原因：
2020-09-18 14:39:07.885039: 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 FMA
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
(2, 2)
2020-09-18 14:39:07.896766: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x7fde58a96d90 initialized for platform Host (this does not guarantee that XLA will be used). Devices:
2020-09-18 14:39:07.896780: I tensorflow/compiler/xla/service/service.cc:176] StreamExecutor device (0): Host, Default Version
tensorflow版本是cpu 2.3.0版本
这不是Warning呀，哪里都没出现warning这个词，只是给你一些information而已。
据我所知第一个warning如果要解决的话，需要从github下载tensorflow的源代码，用bazel重新编译。这个挺复杂的。具体可参考https://tensorflow.google.cn/install/source（我自己因为是使用gpu版本，所以没必要解决这个问题）
第二个warning我还没弄明白，需要好好看看XLA部分的讲解。
你好，请问跑训练代码出现下面两个信息是什么意思？
1、None of the MLIR Optimization Passes are enabled
2、Couldn’t invoke ptxas.exe --version
代码是链接里的示例代码：कनवल्शनल न्यूरल नेटवर्क (सीएनएन) | TensorFlow Core
看起来是能用GPU训练，但是总感觉速度有问题。。跟机器用的是AMD的CPU有关系吗，CPU是5800x，显卡是3080ti，环境是tensorflow-gpu 2.5.0，python=3.8，cudatoolkit=11.3，cudnn=8.2。
具体信息如下：
2021-09-12 17:35:28.302826: I tensorflow/compiler/mlir/mlir_graph_optimization_pass.cc:176] None of the MLIR Optimization Passes are enabled (registered 2)
Epoch 1/10
2021-09-12 17:35:28.528250: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library cudnn64_8.dll
2021-09-12 17:35:28.954801: I tensorflow/stream_executor/cuda/cuda_dnn.cc:359] Loaded cuDNN version 8201
2021-09-12 17:35:29.577930: E tensorflow/core/platform/windows/subprocess.cc:287] Call to CreateProcess failed. Error code: 2
2021-09-12 17:35:29.578036: W tensorflow/stream_executor/gpu/asm_compiler.cc:56] Couldn’t invoke ptxas.exe --version
2021-09-12 17:35:29.582453: E tensorflow/core/platform/windows/subprocess.cc:287] Call to CreateProcess failed. Error code: 2
2021-09-12 17:35:29.582966: W tensorflow/stream_executor/gpu/redzone_allocator.cc:314] Internal: Failed to launch ptxas
Relying on driver to perform ptx compilation.
Modify $PATH to customize ptxas location.
This message will be only logged once.
2021-09-12 17:35:29.620691: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library cublas64_11.dll
2021-09-12 17:35:30.122407: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library cublasLt64_11.dll
2021-09-12 17:35:30.155622: I tensorflow/stream_executor/cuda/cuda_blas.cc:1838] TensorFloat-32 will be used for the matrix multiplication. This will only be logged once.
1563/1563 [==============================] - 7s 3ms/step - loss: 1.5276 - accuracy: 0.4408 - val_loss: 1.2432 - val_accuracy: 0.5580
Epoch 2/10
1563/1563 [==============================] - 4s 3ms/step - loss: 1.1622 - accuracy: 0.5898 - val_loss: 1.1072 - val_accuracy: 0.6117
我也不太确定原因，或许可以参考 python - Tensorflow 2.4.1 - Couldn't invoke ptxas.exe - Stack Overflow ，换个tf和cuda的版本组合看看
您好，想请问按照您的方法和顺序在conda里安装tensorflow和cuda之后，是直接可以使用GPU运行的么？
我按照上面出现了检测不到GPU的情况，并且好像没有缺少什么文件的报错。CPU正常。
print(tf.test.gpu_device_name())有错误信息，输出大概这样。
[] [PhysicalDevice(name=’/physical_device:CPU:0’, device_type=‘CPU’)]
2022-02-18 19:55:37.195693: 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: SSE4.1 SSE4.2 AVX AVX2 FMA
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
版本是
cudatoolkit 11.3.1 h2bc3f7f_2
cudnn 8.2.1 cuda11.3_0
tensorboard 2.6.0 py_1
tensorboard-data-server 0.6.0 py37hca6d32c_0
tensorboard-plugin-wit 1.6.0 py_0
tensorflow 2.6.0 mkl_py37h9d15365_0
tensorflow-base 2.6.0 mkl_py37h3d85931_0
tensorflow-estimator 2.6.0 pyh7b7c402_0
各种重装都试了比较头大，想请问有无这种情况出现过？
1 reply
这里的输出也没报错呀，哪一行说检测不到gpu？