系统在ANN训练中途失败

Question

在进行ANN训练的过程中，代码突然停止或导致BSoD这是代码：

import tensorflow as tf
from tensorflow import keras
from keras.models import Sequential
from keras.layers import Dense,Flatten
from keras.optimizers import adam
from keras.activations import relu,softmax

dataset = keras.datasets.fashion_mnist
(train_images,train_labels),(test_images,test_labels) = dataset.load_data()

train_images = train_images / 255.0
test_images = test_images / 255.0
class_names = [
"T-shirt/top",
"Trouser",
"Pullover",
"Dress",
"Coat",
"Sandal",
"Shirt",
"Sneaker",
"Bag",
"Ankle boot"
]

model = Sequential()
model.add(Flatten(input_shape=(28,28)))
model.add(Dense(128,activation="relu"))
model.add(Dense(10,activation="softmax"))

model.compile(optimizer="adam",loss="sparse_categorical_crossentropy",metrics=["accuracy"])

model.fit(train_images,train_labels,epochs=20,batch_size=12)

test_loss, test_acc = model.evaluate(test_images,test_labels)

print(f"test accuracy: {test_acc}")
print(f"test loss: {test_loss}")

这里是不引起BSoD时的全部错误消息：

C:\Users\User\AppData\Local\Programs\Python\Python37\python.exe C:/Users/User/PycharmProjects/AI/tutorial.py
2020-01-26 17:53:22.098836: W tensorflow/stream_executor/platform/default/dso_loader.cc:55] Could not load dynamic library 'cudart64_101.dll'; dlerror: cudart64_101.dll not found
2020-01-26 17:53:22.099178: 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.
Using TensorFlow backend.
2020-01-26 17:53:27.826205: W tensorflow/stream_executor/platform/default/dso_loader.cc:55] Could not load dynamic library 'nvcuda.dll'; dlerror: nvcuda.dll not found
2020-01-26 17:53:27.826375: E tensorflow/stream_executor/cuda/cuda_driver.cc:351] failed call to cuInit: UNKNOWN ERROR (303)
2020-01-26 17:53:27.828951: I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:169] retrieving CUDA diagnostic information for host: DESKTOP-507H6IM
2020-01-26 17:53:27.829139: I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:176] hostname: DESKTOP-507H6IM
2020-01-26 17:53:27.831554: I tensorflow/core/platform/cpu_feature_guard.cc:142] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2
Epoch 1/20

   12/60000 [..............................] - ETA: 13:55 - loss: 2.5218 - accuracy: 0.1667
  600/60000 [..............................] - ETA: 21s - loss: 1.3435 - accuracy: 0.5400  
 1200/60000 [..............................] - ETA: 13s - loss: 1.1029 - accuracy: 0.6200
 1812/60000 [..............................] - ETA: 10s - loss: 0.9730 - accuracy: 0.6634
 2424/60000 [>.............................] - ETA: 8s - loss: 0.9054 - accuracy: 0.6852 
 3048/60000 [>.............................] - ETA: 7s - loss: 0.8507 - accuracy: 0.7008
 3636/60000 [>.............................] - ETA: 7s - loss: 0.8226 - accuracy: 0.7109
 4200/60000 [=>............................] - ETA: 6s - loss: 0.7984 - accuracy: 0.7176
 4836/60000 [=>............................] - ETA: 6s - loss: 0.7760 - accuracy: 0.7235
 5460/60000 [=>............................] - ETA: 6s - loss: 0.7497 - accuracy: 0.7313
 6060/60000 [==>...........................] - ETA: 5s - loss: 0.7253 - accuracy: 0.7421
 6636/60000 [==>...........................] - ETA: 6s - loss: 0.7191 - accuracy: 0.7455
 7056/60000 [==>...........................] - ETA: 5:52 - loss: 0.7101 - accuracy: 0.7476
 7440/60000 [==>...........................] - ETA: 5:31 - loss: 0.6963 - accuracy: 0.7519


Process finished with exit code -1073741819 (0xC0000005)

它可能与Tensorflow构建有关系吗？我使用pip安装了它如果我应该使用Basel构建自己的Tensorflow，请告诉我如何操作。

我正在使用pycharm 2019.3

Windows 10

python 3.7

Tensorflow 2.0.0

CPU：AMD锐龙5 3600

Answer 1

您的系统完全有能力执行工作负载。（您是否超频了CPU？这可能也会导致不稳定。）

只需卸载tensorflow，python，也许重置pycharm（导出首选项）。并重新安装软件包。这对我来说很好。通常，这可以解决。

但是我有一个对此不起作用的朋友。我们必须重置他的Windows才能解决此问题。

希望这会有所帮助。