心得：此代码可以很好地体现残差网络的构建
The complexity and accuracy of the neural network model
Dependencies:
- Windows10
- python==3.6.12
-
GeForce GTX 1660TI
- tensorflow-gpu==2.0.0
- GPU环境安装包，下载地址：https://pan.baidu.com/s/14Oisbo9cZpP7INQ6T-3vwA 提取码：z4pl （网上找的）
Anaconda3-5.2.0-Windows-x86_64.exe
cuda_10.0.130_411.31_win10.exe
cudnn-10.0-windows10-x64-v7.4.2.24.zip
h5py-2.8.0rc1-cp36-cp36m-win_amd64.whl
numpy-1.16.4-cp36-cp36m-win_amd64.whl
tensorflow_gpu-1.13.1-cp36-cp36m-win_amd64.whl
torch-1.1.0-cp36-cp36m-win_amd64.whl
torchvision-0.3.0-cp36-cp36m-win_amd64.whl
Visualization Results
- resnet18 训练结果 1
- resnet101 训练结果 2
Public Datasets:
- cifar100
Experience：
Traceback (most recent call last):
File "resnet18_train.py", line 89, in <module>
main()
File "resnet18_train.py", line 53, in main
logits = model(x)
File "D:\installpackage\ana\envs\tf361\lib\site-packages\tensorflow_core\python\keras\engine\base_layer.py", line 891, in __call__
outputs = self.call(cast_inputs, *args, **kwargs)
File "C:\Users\33203\Desktop\lesson43\resnet.py", line 72, in call
x = self.stem(inputs)
File "D:\installpackage\ana\envs\tf361\lib\site-packages\tensorflow_core\python\keras\engine\base_layer.py", line 891, in __call__
outputs = self.call(cast_inputs, *args, **kwargs)
File "D:\installpackage\ana\envs\tf361\lib\site-packages\tensorflow_core\python\keras\engine\sequential.py", line 270, in call
outputs = layer(inputs, **kwargs)
File "D:\installpackage\ana\envs\tf361\lib\site-packages\tensorflow_core\python\keras\engine\base_layer.py", line 891, in __call__
outputs = self.call(cast_inputs, *args, **kwargs)
File "D:\installpackage\ana\envs\tf361\lib\site-packages\tensorflow_core\python\keras\layers\convolutional.py", line 197, in call
outputs = self._convolution_op(inputs, self.kernel)
File "D:\installpackage\ana\envs\tf361\lib\site-packages\tensorflow_core\python\ops\nn_ops.py", line 1134, in __call__
return self.conv_op(inp, filter)
File "D:\installpackage\ana\envs\tf361\lib\site-packages\tensorflow_core\python\ops\nn_ops.py", line 639, in __call__
return self.call(inp, filter)
File "D:\installpackage\ana\envs\tf361\lib\site-packages\tensorflow_core\python\ops\nn_ops.py", line 238, in __call__
name=self.name)
File "D:\installpackage\ana\envs\tf361\lib\site-packages\tensorflow_core\python\ops\nn_ops.py", line 2010, in conv2d
name=name)
File "D:\installpackage\ana\envs\tf361\lib\site-packages\tensorflow_core\python\ops\gen_nn_ops.py", line 1031, in conv2d
data_format=data_format, dilations=dilations, name=name, ctx=_ctx)
File "D:\installpackage\ana\envs\tf361\lib\site-packages\tensorflow_core\python\ops\gen_nn_ops.py", line 1130, in conv2d_eager_fallback
ctx=_ctx, name=name)
File "D:\installpackage\ana\envs\tf361\lib\site-packages\tensorflow_core\python\eager\execute.py", line 67, in quick_execute
six.raise_from(core._status_to_exception(e.code, message), None)
File "<string>", line 3, in raise_from
tensorflow.python.framework.errors_impl.UnknownError: Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so try looking to see if a warning log message was printed above. [Op:Conv2D]
出现上面的错误
- 原因一：You're out of memory
- reducing your batch size
- using a simpler model
- using less data
- limit TensorFlow GPU memory（本代码按此方法解决）： https://stackoverflow.com/questions/53698035/failed-to-get-convolution-algorithm-this-is-probably-because-cudnn-failed-to-in(参考此网址)
os.environ['TF_FORCE_GPU_ALLOW_GROWTH'] = 'true'**-----------------------------------（本代码按此方法解决）
or
physical_devices = tf.config.experimental.list_physical_devices('GPU')
if len(physical_devices) > 0:
tf.config.experimental.set_memory_growth(physical_devices[0], True)
- 原因二：You have incompatible versions of CUDA, TensorFlow, NVIDIA drivers, etc.
- you can see https://blog.csdn.net/qq_41683065/article/details/108702408 my cudnn==7.6.4 cuda10.0_0 cudatoolkit==10.0.130
tensorflow-gpu版本代码出现numpy错误
- 其中一种解决方法：pip install --upgrade numpy
References:
- 深度学习与TensorFlow 2入门实战（完整版）---龙曲良