Python 相同的Keras脚本在WSL（Ubuntu）中工作，但在Conda中不工作？

我正在训练CNN对图像进行分类，我有一个Keras脚本，它可以在CPU和GPU环境中成功运行，但是网络只能在CPU环境中学习。在GPU环境中，损耗在第一个历元之后减小，但在第一个历元之后保持不变。为什么?

使用Python3.6，我有一个运行在WSL（Ubuntu）中的tensorflow环境和一个运行在Conda中的tensorflow gpu环境。我尝试过不同的体系结构和不同的优化器，但不管GPU环境有什么问题

编辑：我创建了一个CPU conda环境，我也有同样的问题，所以这似乎是conda与WSL的问题，而不是GPU与CPU的问题。此外，Conda在CPU上的时间周期是WSL中CPU时间周期的两倍

我注释掉了model.fit（），以减少冗长的输出

将numpy导入为np
导入tensorflow作为tf
从tensorflow.keras.layers导入输入、Conv2D、展平、密集、MaxPoolig2D
从tensorflow.keras.models导入模型
导入UTIL
（x_列，y_列），，（x_测试，y_测试）=最终负载数据（限值=36）
输入\图像=输入（形状=（256，256，3））
x=Conv2D（32，（3，3），padding='same'，activation='relu'）（输入图像）
x=Conv2D（32，（3，3），activation='relu'）（x）
x=MaxPoolig2D（）（x）
x=展平（）（x）
x=密集（128，激活='relu'）（x）
输出=密集（9，激活='softmax'）（x）
模型=模型（输入=输入\图像，输出=输出）
compile（loss='classifical_crossentropy'，optimizer='rmsprop'，metrics=['acc']）
#模型拟合（x_序列，y_序列，历元=10，验证数据=（x_检验，y_检验））
对于范围（25）内的i：
打印（str（i）+'：'+str（批量上的型号列（x列，y列）））
模型评估（x_列车、y_列车）
模型评估（x_检验，y_检验）

CPU培训结果：

0: [17.486359, 0.6875]
1: [61761.203, 0.28125]
2: [2228.4707, 0.71875]
3: [4440.558, 0.28125]
4: [1062.5581, 0.71875]
5: [481.29315, 0.71875]
6: [234.01581, 0.4375]
7: [170.98215, 0.71875]
8: [38.968575, 0.6875]
9: [8.086919, 0.75]
10: [5.7502546, 0.375]
11: [72.89319, 0.71875]
12: [13.203195, 0.6875]
13: [1.4184309, 0.875]
14: [9.258236, 0.46875]
15: [23.165062, 0.71875]
16: [8.963888, 0.78125]
17: [3.1053305, 0.84375]
18: [1.0664859, 0.96875]
19: [0.039992813, 1.0]
20: [0.023323938, 1.0]
21: [0.019487603, 1.0]
22: [0.01734325, 1.0]
23: [0.015670585, 1.0]
24: [0.014209943, 1.0]
32/32 [==============================] - 1s 19ms/sample - loss: 0.0129 - acc: 1.0000
4/4 [==============================] - 0s 20ms/sample - loss: 2.3463 - acc: 0.7500

0: [8.630159, 0.1875]
1: [4.5332146, 0.71875]
2: [4.5332146, 0.71875]
3: [4.5332146, 0.71875]
4: [4.5332146, 0.71875]
5: [4.5332146, 0.71875]
6: [4.5332146, 0.71875]
7: [4.5332146, 0.71875]
8: [4.5332146, 0.71875]
9: [4.5332146, 0.71875]
10: [4.5332146, 0.71875]
11: [4.5332146, 0.71875]
12: [4.5332146, 0.71875]
13: [4.5332146, 0.71875]
14: [4.5332146, 0.71875]
15: [4.5332146, 0.71875]
16: [4.5332146, 0.71875]
17: [4.5332146, 0.71875]
18: [4.5332146, 0.71875]
19: [4.5332146, 0.71875]
20: [4.5332146, 0.71875]
21: [4.5332146, 0.71875]
22: [4.5332146, 0.71875]
23: [4.5332146, 0.71875]
24: [4.5332146, 0.71875]
32/32 [==============================] - 0s 4ms/sample - loss: 4.5332 - acc: 0.7188
4/4 [==============================] - 0s 19ms/sample - loss: 4.0295 - acc: 0.7500

我希望看到与上述类似的东西，但我得到了一个奇怪的东西——GPU培训的结果：

0: [17.486359, 0.6875]
1: [61761.203, 0.28125]
2: [2228.4707, 0.71875]
3: [4440.558, 0.28125]
4: [1062.5581, 0.71875]
5: [481.29315, 0.71875]
6: [234.01581, 0.4375]
7: [170.98215, 0.71875]
8: [38.968575, 0.6875]
9: [8.086919, 0.75]
10: [5.7502546, 0.375]
11: [72.89319, 0.71875]
12: [13.203195, 0.6875]
13: [1.4184309, 0.875]
14: [9.258236, 0.46875]
15: [23.165062, 0.71875]
16: [8.963888, 0.78125]
17: [3.1053305, 0.84375]
18: [1.0664859, 0.96875]
19: [0.039992813, 1.0]
20: [0.023323938, 1.0]
21: [0.019487603, 1.0]
22: [0.01734325, 1.0]
23: [0.015670585, 1.0]
24: [0.014209943, 1.0]
32/32 [==============================] - 1s 19ms/sample - loss: 0.0129 - acc: 1.0000
4/4 [==============================] - 0s 20ms/sample - loss: 2.3463 - acc: 0.7500

0: [8.630159, 0.1875]
1: [4.5332146, 0.71875]
2: [4.5332146, 0.71875]
3: [4.5332146, 0.71875]
4: [4.5332146, 0.71875]
5: [4.5332146, 0.71875]
6: [4.5332146, 0.71875]
7: [4.5332146, 0.71875]
8: [4.5332146, 0.71875]
9: [4.5332146, 0.71875]
10: [4.5332146, 0.71875]
11: [4.5332146, 0.71875]
12: [4.5332146, 0.71875]
13: [4.5332146, 0.71875]
14: [4.5332146, 0.71875]
15: [4.5332146, 0.71875]
16: [4.5332146, 0.71875]
17: [4.5332146, 0.71875]
18: [4.5332146, 0.71875]
19: [4.5332146, 0.71875]
20: [4.5332146, 0.71875]
21: [4.5332146, 0.71875]
22: [4.5332146, 0.71875]
23: [4.5332146, 0.71875]
24: [4.5332146, 0.71875]
32/32 [==============================] - 0s 4ms/sample - loss: 4.5332 - acc: 0.7188
4/4 [==============================] - 0s 19ms/sample - loss: 4.0295 - acc: 0.7500

---

我迫不及待地想看看我犯了什么愚蠢的错误，哈哈，我不知道实际的问题是什么，但我知道从1.13.1更新tensorflow gpu修复了它

---

在conda shell中，我运行了

pip install tf nightly gpu

，现在网络按预期运行。我打赌我不必使用夜间构建，我可以只指定1.14.0（我在WSL中使用的tensorflow gpu构建），但不管怎样。

我希望我可以复制它。你有完全相同的代码和数据吗？相同的32个数据点？同样的分裂？（数据集是什么？@Mehdi完全相同的代码和数据。不过，我刚刚意识到的一点是，当使用CPU conda环境时，我会遇到同样的问题。所以这看起来像是康达vs WSL的问题，而不是CPU vs GPU的问题。@Mehdi这是ISIC挑战赛的训练集。我还注意到康达的训练速度是WSL的两倍。。。