TensorFlow是否支持M1 max？

Question

在我尝试在我的mac蒙特利12.6.1上开始使用TensorFlow时，我开始发现在我的mac迷你蒙特利12.6芯片M1 2020上没有观察到的错误

这要么是环境问题，要么是芯片组问题。在我的windows机器Win-11和Mac-Mini上工作。

代码：

from tensorflow.keras.models import Sequential
from tensorflow.keras.optimizers import Adam
from tensorflow.keras import layers

model = Sequential([layers.Input((3, 1)),
                    layers.LSTM(64),
                    layers.Dense(32, activation='relu'),
                    layers.Dense(32, activation='relu'),
                    layers.Dense(1)])

model.compile(loss='mse', 
              optimizer=Adam(learning_rate=0.001),
              metrics=['mean_absolute_error'])

model.fit(X_train, y_train, validation_data=(X_val, y_val), epochs=100)

在DataSpell中观察到的错误：

--------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
RuntimeError: module compiled against API version 0x10 but this version of numpy is 0xe

回溯

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Input In [14], in <cell line: 1>()
----> 1 from tensorflow.keras.models import Sequential
      2 from tensorflow.keras.optimizers import Adam
      3 from tensorflow.keras import layers

下面是Greg教程：https://www.youtube.com/watch?v=CbTU92pbDKw

注意，这段代码工作在我的machine机器上，而不是在MacBook Pro上。Anaconda env -> Python3.9

python --version
Python 3.9.12

conda list | grep tensorflow
tensorflow-deps           2.8.0                         0    apple
tensorflow-estimator      2.10.0                   pypi_0    pypi
tensorflow-macos          2.10.0                   pypi_0    pypi
tensorflow-metal          0.6.0                    pypi_0    pypi

我期待的是类似的结果，如windows环境和Mac-mini，其中的模型是构建和安装的培训数据。(毫无例外地创建模型对象)

示例：

Epoch 99/100
7/7 [==============================] - 0s 5ms/step - loss: 6.1541 - mean_absolute_error: 1.8648 - val_loss: 9.5456 - val_mean_absolute_error: 2.6235
Epoch 100/100
7/7 [==============================] - 0s 5ms/step - loss: 6.7555 - mean_absolute_error: 2.0134 - val_loss: 9.4403 - val_mean_absolute_error: 2.6016

<keras.callbacks.History at 0x27a6590c6a0>

在尝试numpy升级时，我做了"numpy升级“，但是终端上仍然有下面的输出，同样的异常仍然被观察到。

pip install numpy --upgrade
Requirement already satisfied: numpy in ./opt/anaconda3/lib/python3.9/site-packages (1.21.5)
Collecting numpy
  Using cached numpy-1.23.4-cp39-cp39-macosx_11_0_arm64.whl (13.4 MB)
Installing collected packages: numpy
  Attempting uninstall: numpy
    Found existing installation: numpy 1.21.5
    Uninstalling numpy-1.21.5:
      Successfully uninstalled numpy-1.21.5
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
scipy 1.7.3 requires numpy<1.23.0,>=1.16.5, but you have numpy 1.23.4 which is incompatible.
numba 0.55.1 requires numpy<1.22,>=1.18, but you have numpy 1.23.4 which is incompatible.
Successfully installed numpy-1.23.4

==================================

因此，多种方法的结合解决了这一问题：

1. pip卸载角
2. pip卸载keras-预处理
3. pip卸载张紧板
4. pip安装--升级numpy
5. 如果步骤4不工作错误或有关警告，则pip卸载numpy；然后是pip install numpy。
6. python安装tensorflow-macos来解决我的环境问题。

Answer 1

从苹果文档到tensorflow安装

要求

• 带有苹果硅或AMD GPU的Mac电脑
• macOS 12.0或更高版本(获取最新测试版)
• Python 3.8或更高版本
• Xcode命令行工具: xcode-select --安装

设置

// Download Conda environment
bash ~/miniconda.sh -b -p $HOME/miniconda
source ~/miniconda/bin/activate
conda install -c apple tensorflow-deps

// Install base TensorFlow
python -m pip install tensorflow-macos

// Install tensorflow-metal plug-in
python -m pip install tensorflow-metal

验证设置

import tensorflow as tf

cifar = tf.keras.datasets.cifar100
(x_train, y_train), (x_test, y_test) = cifar.load_data()
model = tf.keras.applications.ResNet50(
    include_top=True,
    weights=None,
    input_shape=(32, 32, 3),
    classes=100,)

loss_fn = tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True)
model.compile(optimizer="adam", loss=loss_fn, metrics=["accuracy"])
model.fit(x_train, y_train, epochs=5, batch_size=64)

Answer 2

但是，解决问题的真正答案是，您安装了许多版本的numpy。

请在您的终端上运行pip3 install --upgrade numpy，它会解决您的问题。

Answer 3

因此，多种方法的结合解决了这一问题：

1. pip uninstall keras
2. pip uninstall keras-preprocessing
3. pip uninstall tensorboard
4. pip install --upgrade numpy
5. 如果步骤4不工作错误或有关警告，那么pip uninstall numpy；后面跟着pip install numpy
6. python -m pip install tensorflow-macos

解决了我的环境问题。