Problem Description:
an error is reported when installing tensorflow GPU = = 1.4.1 in the virtual environment with Python version 3.6.0
AttAttributeError: module ‘typing’ has no attribute ‘NoReturn’
Cause analysis:
Python version is too low. Python has changed a lot since 3.6
Solution:
install a newer version of python, and you can solve the problem above 3.6.2
Running with tensorflow 1.15, the following error occurs:,
terms of settlement:
It is found that the tensorflow estimator installed automatically when installing 1.15.0 is version 2.5.0 (Figure 2)
Uninstall version 2.5: PIP uninstall tensorflow-estimator
Reinstall version 1.15: CONDA install tensorflow-estimator==1.15.1 Or pip install Tensorflow estimator = = 1.15.1 (I installed it with Canda, but failed to use pip, prompting an error, valueerror: check_ hostname requires server_ Hostname, after searching this error, I need to close the agent, but I rely on the agent to connect to the network but not close…)
Problem solving.
Note: the reason for installing 1.15.1 is that version 1.15.0 is not found. If you want to download the manual installation from the official website, you can only find tensorflow CPU estimator 1.15.1, search results · pypi
Reference: modulenotfounderror: no module named ‘tensorflow. Python. Types’ – stack overflow
CUDA driver version is insufficient for CUDA runtime version
Question:
An error is reported when docker runs ONEFLOW code of insightface
Failed to get cuda runtime version: CUDA driver version is insufficient for CUDA runtime version
reason:
1. View CUDA runtime version
cat /usr/local/cuda/version.txt
The CUDA version in my docker is 10.0.130
CUDA Version 10.0.130
2. The CUDA version has requirements for the graphics card driver version, see the following link.
https://docs.nvidia.com/cuda/cuda-toolkit-release-notes/index.html
| CUDA Toolkit | Linux x86 64 Driver Version | Windows x86 and 64 Driver Version |
|---|---|---|
| CUDA 11.0.3 Update 1 | ||
| CUDA 11.0.2 GA | >= 450.51.05 | >= 451.48 |
| CUDA 11.0.1 RC | >= 450.36.06 | >= 451.22 |
| CUDA 10.2.89 | >= 440.33 | >= 441.22 |
| CUDA 10.1 (10.1.105 general release, and updates) | >= 418.39 | >= 418.96 |
| CUDA 10.0.130 | >= 410.48 | >= 411.31 |
| CUDA 9.2 (9.2.148 Update 1) | >= 396.37 | >= 398.26 |
| CUDA 9.2 (9.2.88) | >= 396.26 | >= 397.44 |
cat /proc/driver/nvidia/version took a look at the server’s graphics card driver is 418.67, CUDA 10.1 should be installed, and I installed 10.0.130 cuda.
NVRM version: NVIDIA UNIX x86_64 Kernel Module 418.67 Sat Apr 6 03:07:24 CDT 2019
GCC version: gcc version 7.3.0 (Ubuntu 7.3.0-27ubuntu1~18.04)
solve:
Installing CUDA 10.1
(1) First in https://developer.nvidia.com/cuda-toolkit-archive According to the machine environment, download the corresponding cuda10.1 installation file. For the installer type, I choose runfile (local). The installation steps will be simpler.
wget https://developer.download.nvidia.com/compute/cuda/10.1/Prod/local_installers/cuda_10.1.243_418.87.00_linux.runsudo sh
(2) Installation
sh cuda_10.1.243_418.87.00_linux.run
The same error occurred, unresolved
it will be updated when a solution is found later.
The first time you use tensorflow’s debug tool, but when you use it for the second time, there is always a shortage of space, which can be solved through the following steps.
df -h
Find that the root directory is full, then go to the root directory and check the occupied directory
du –max-depth=1 -h
It is found that TMP directory takes up a lot of space
Sure enough, when you go to TMP, you find files related to tfdbg. Just delete them 。
Type error: cannot unpack non Iterable non type object appears when starting Bert server in Ubuntu system
Questions
Enter Bert serving start – model_ dir chinese_ L-12_ H-768_ A-12 -num_ worker 1 -max_ seq_ In the case of len 64, type error: cannot unpack non Iterable non type object is reported
resolvent
Finally, it can run without any error
Contents of articles
Problem description solution references
Problem description
When you pull the model trained by the server to your own computer, tensorflow loads the model and reports an error valueerror: unknown layer: functional error
import tensorflow as tf
model = tf.keras.models.load_model('test.h5')
The server
python 3.6.13
tensorflow-gpu==2.3.0
Own computer
python 3.6.5
tensorflow-gpu==2.1.0
Solution
pip install tensorflow-gpu==2.3.0
References
Server environment:
Recently, when I was running demo pointasnl of point cloud classification, when batch_ When the size setting is relatively large, the following errors will appear during the training:
2020-06-12 00:14:01.824110: E tensorflow/stream_executor/cuda/cuda_event.cc:29] Error polling for event status: failed to query event: CUDA_ERROR_ILLEGAL_ADDRESS: an illegal memory access was encountered
2020-06-12 00:14:01.824142: F tensorflow/core/common_runtime/gpu/gpu_event_mgr.cc:273] Unexpected Event status: 1At first, it was thought that there was something wrong with the GPU Programming code, but after repeated checking, it was found that there was no error.
After collecting information from the Internet, I vaguely realized that it should be the environmental version.
After reducing cudnn 7.4.5 to cudnn 7.3.1, this problem seems to be solved. I hope there will be no more problems.
Project scenario:
under Windows environment, python 3.6, the versions of each CONDA package are as follows 0
\# Name Version Build Channel
absl-py 0.13.0 <pip>
astor 0.8.1 <pip>
cached-property 1.5.2 <pip>
certifi 2021.5.30 py36ha15d459_0 https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
cycler 0.10.0 <pip>
dataclasses 0.8 <pip>
gast 0.2.2 <pip>
google-pasta 0.2.0 <pip>
grpcio 1.38.1 <pip>
h5py 3.1.0 <pip>
importlib-metadata 4.6.0 <pip>
joblib 1.0.1 <pip>
Keras 2.3.1 <pip>
Keras-Applications 1.0.8 <pip>
Keras-Preprocessing 1.1.2 <pip>
kiwisolver 1.3.1 <pip>
Markdown 3.3.4 <pip>
matplotlib 3.3.4 <pip>
numpy 1.19.5 <pip>
opt-einsum 3.3.0 <pip>
pandas 1.1.5 <pip>
Pillow 8.2.0 <pip>
pip 21.1.3 pyhd8ed1ab_0 https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
protobuf 3.17.3 <pip>
pyparsing 2.4.7 <pip>
python 3.6.13 h39d44d4_0_cpython https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
python-dateutil 2.8.1 <pip>
python_abi 3.6 2_cp36m https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
pytz 2021.1 <pip>
PyYAML 5.4.1 <pip>
scikit-learn 0.24.2 <pip>
scipy 1.5.4 <pip>
seaborn 0.11.1 <pip>
setuptools 49.6.0 py36ha15d459_3 https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
six 1.16.0 <pip>
sklearn 0.0 <pip>
tensorboard 1.15.0 <pip>
tensorflow 1.15.0 <pip>
tensorflow-estimator 1.15.1 <pip>
termcolor 1.1.0 <pip>
threadpoolctl 2.1.0 <pip>
typing-extensions 3.10.0.0 <pip>
ucrt 10.0.20348.0 h57928b3_0 https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
vc 14.2 hb210afc_5 https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
vs2015_runtime 14.29.30037 h902a5da_5 https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
Werkzeug 2.0.1 <pip>
wheel 0.36.2 pyhd3deb0d_0 https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
wincertstore 0.2 py36ha15d459_1006 https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
wrapt 1.12.1 <pip>
zipp 3.4.1 <pip>
Problem Description:
it seems that the version of the library that has been adjusted for a long time has finally run through a LSTM demo, but if you want to open a new test file for testing, the following problems appear
Solution:
the solution to stackoverflow is to use the current keras library and tensorflow Library of unitall, because the higher version of keras library may be used, but the running code is of lower version. Note that you need to install tensorflow before installing keras
The specific steps of stackoverflow with the most likes are as follows:
delete tenorflow delete tenorflow
pip uninstall tensorflow
Update PIP version
pip install --upgrade pip
install keras
pip install Keras
install tensorflow
pip install tensorflow
Apart from this method, I want to say that my final solution is: I found the import statement of the keras library
from keras.models import Sequential
from keras.layers import LSTM, Dense, Activation
Cannot appear in two. Py files!!! Even if you just don’t use import, you will still report this error. It’s too bad. I found this problem after reloading it, so when testing the code, you should either create a new project or write the import statement in a file.
After the installation of tensorflow2. X is successful, after running the following code:
tf.config.list_physical_devices('GPU')There are always the following situations: (Note: there are usually multiple, only two here are shown here)
Could not load dynamic library ‘cublas64_ 10.dll’; dlerror: cublas64_ 10.dll not found
Could not load dynamic library ‘cudnn64_ 7.dll’; dlerror: cudnn64_ 7.dll not found
Solution:
Download the corresponding DLL file and put it into the folder C: Windows: system32.
The problem was solved successfully Pro test (valid)
So the key to the problem, where to download these DLL files, rest assured, this article will provide you with all the required DLL files.
Error information:
Traceback (most recent call last):
File "bs.py", line 149, in <module>
tcpserver1=MYTCPServer(('192.168.0.109',54321))
File "wserver_bs.py", line 65, in __init__
self.model.load_weights(weight_filepath)
File "/home/ps/anaconda3/envs/anomaly/lib/python3.6/site-packages/tensorflow/python/keras/engine/training.py", line 162, in load_weights
return super(Model, self).load_weights(filepath, by_name)
File "/home/ps/anaconda3/envs/anomaly/lib/python3.6/site-packages/tensorflow/python/keras/engine/network.py", line 1424, in load_weights
saving.load_weights_from_hdf5_group(f, self.layers)
File "/home/ps/anaconda3/envs/anomaly/lib/python3.6/site-packages/tensorflow/python/keras/saving/hdf5_format.py", line 749, in load_weights_from_hdf5_group
layer, weight_values, original_keras_version, original_backend)
File "/home/ps/anaconda3/envs/anomaly/lib/python3.6/site-packages/tensorflow/python/keras/saving/hdf5_format.py", line 456, in preprocess_weights_for_loading
weights[0] = np.transpose(weights[0], (3, 2, 0, 1))
File "<__array_function__ internals>", line 6, in transpose
File "/home/ps/anaconda3/envs/anomaly/lib/python3.6/site-packages/numpy/core/fromnumeric.py", line 653, in transpose
return _wrapfunc(a, 'transpose', axes)
File "/home/ps/anaconda3/envs/anomaly/lib/python3.6/site-packages/numpy/core/fromnumeric.py", line 58, in _wrapfunc
return bound(*args, **kwds)
ValueError: axes don't match array
Tossed half a night, the Internet to check a variety of methods are useless
the fault lies in load_ On the weights function, it’s a matter of setting the model!
When loading the model, the default input size is used
model = cnn.CNNLikeModel()
The actual size of the input tensor is different from the default size, which leads to this error.
solve
1. Modify the default value of the function
2. Fill in the correct value when using this function.
Dlerror: libcudart. So. 11.0: problem solving
First find your computer path
/usr/local/cuda/lib64
Check your CUDA version
. According to the above figure, I find that my computer’s CUDA version is 10.0, so I will report an error at runtime. At this time, there are two solutions.
Scheme 1
If you force the CUDA environment of the local computer to be the running CUDA environment, there may be problems, which I haven’t tried.
cd /usr/local/cuda/lib64/
sudo ln -sf libcudart.so.10.0 libcudart.so.11.0
Scheme 2
Installing dynamic CUDA environment in CONDA environment
conda install tensorflow-gpu cudatoolkit=11.0
**
TypeError: super(type, obj): obj must be an instance or subtype of type
**
Solution:
Just close the IDE and restart it.