乐趣区

关于python:金玉良缘易配而木石前盟难得M1-Mac-配置Python3开发环境搭建集成TensorflowPytorch

原文转载自「刘悦的技术博客」https://v3u.cn/a_id_189

笔者投入 M1 的怀抱曾经有一段时间了,俗话说得好,但闻新人笑,不见旧人哭,Intel mac 早已被束之高阁,而 M1 mac 曾经不能用真香来形容了,几乎就是“香透满堂金玉彩,扇遮半面桃花开!”,轻抚 M1 mac 那滑若柔荑的秒控键盘,别说 996 了,就是 007,咱们也能安之若素,也能够笑慰平生。好了,日常吹 M1 的环节完结,正所谓剑虽利, 不厉一直,材虽美,不学不高。本次咱们尝试在 M1 Mac os 中搭建 Python3 的开发环境。

个别状况下,间接 Python 官网 (python.org) 下载最新的基于 arm 架构的 python3.9 即可,然而因为向下兼容等问题,咱们尝试应用 Python 多版本管理软件 conda,conda 在业界有三大巨头别离是:Anaconda、Miniconda 以及 Condaforge,尽管都放出音讯要适配 M1 芯片,然而目前最先放出稳定版的是 Condaforge,进入下载页面:https://github.com/conda-forg… 抉择 mac arm64 位架构:

该文件并不是安装包,而是一个 shell 脚本,下载胜利后,进入命令行目录:

cd ~/Downloas

执行命令进行装置:

sudo bash ./Miniforge3-MacOSX-arm64.sh

随后会有一些条款须要确认,这里按回车之后键入 yes:

Welcome to Miniforge3 4.9.2-7  
  
In order to continue the installation process, please review the license  
agreement.  
Please, press ENTER to continue  
>>>   
BSD 3-clause license  
Copyright (c) 2019-2020, conda-forge  
All rights reserved.  
  
Redistribution and use in source and binary forms, with or without  
modification, are permitted provided that the following conditions are met:  
  
1. Redistributions of source code must retain the above copyright notice, this  
list of conditions and the following disclaimer.  
  
2. Redistributions in binary form must reproduce the above copyright notice,  
this list of conditions and the following disclaimer in the documentation  
and/or other materials provided with the distribution.  
  
3. Neither the name of the copyright holder nor the names of its contributors  
may be used to endorse or promote products derived from this software without  
specific prior written permission.  
  
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND  
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED  
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE  
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE  
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL  
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR  
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER  
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,  
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE  
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.  
  
  
Do you accept the license terms? [yes|no]  
[no] >>> yes

装置的默认版本还是 3.9,会附带装置 35 个根底库,这样就不必咱们本人手动装置了:

brotlipy                    0.7.0  py39h46acfd9_1001   installed           
  bzip2                       1.0.8  h27ca646_4          installed           
  ca-certificates         2020.12.5  h4653dfc_0          installed           
  certifi                 2020.12.5  py39h2804cbe_1      installed           
  cffi                       1.14.5  py39h702c04f_0      installed           
  chardet                     4.0.0  py39h2804cbe_1      installed           
  conda                       4.9.2  py39h2804cbe_0      installed           
  conda-package-handling      1.7.2  py39h51e6412_0      installed           
  cryptography                3.4.4  py39h6e07874_0      installed           
  idna                         2.10  pyh9f0ad1d_0        installed           
  libcxx                     11.0.1  h168391b_0          installed           
  libffi                        3.3  h9f76cd9_2          installed           
  ncurses                       6.2  h9aa5885_4          installed           
  openssl                    1.1.1j  h27ca646_0          installed           
  pip                        21.0.1  pyhd8ed1ab_0        installed           
  pycosat                     0.6.3  py39h46acfd9_1006   installed           
  pycparser                    2.20  pyh9f0ad1d_2        installed           
  pyopenssl                  20.0.1  pyhd8ed1ab_0        installed           
  pysocks                     1.7.1  py39h2804cbe_3      installed           
  python                      3.9.2  hcbd9b3a_0_cpython  installed           
  python_abi                    3.9  1_cp39              installed           
  readline                      8.0  hc8eb9b7_2          installed           
  requests                   2.25.1  pyhd3deb0d_0        installed           
  ruamel_yaml               0.15.80  py39h46acfd9_1004   installed           
  setuptools                 49.6.0  py39h2804cbe_3      installed           
  six                        1.15.0  pyh9f0ad1d_0        installed           
  sqlite                     3.34.0  h6d56c25_0          installed           
  tk                         8.6.10  hf7e6567_1          installed           
  tqdm                       4.57.0  pyhd8ed1ab_0        installed           
  tzdata                      2021a  he74cb21_0          installed           
  urllib3                    1.26.3  pyhd8ed1ab_0        installed           
  wheel                      0.36.2  pyhd3deb0d_0        installed           
  xz                          5.2.5  h642e427_1          installed           
  yaml                        0.2.5  h642e427_0          installed           
  zlib                       1.2.11  h31e879b_1009       installed

而后编辑配置文件 vim ~/.zshrc,退出如下内容(此处 liuyue 是笔者用户名,需改成你的 Mac 以后用户名):

path=('/Users/liuyue/miniforge3/bin' $path)  
export PATH

存盘之后执行命令:

source ~/.zshrc

配置好环境变量之后,键入 python3:

➜  ~ python3  
Python 3.9.2 | packaged by conda-forge | (default, Feb 21 2021, 05:00:30)   
[Clang 11.0.1] on darwin  
Type "help", "copyright", "credits" or "license" for more information.  
>>>

能够看到曾经应用 conda 装置的 python 版本了。

这里简略介绍一下 conda 命令:

conda info 能够查看以后 conda 的根本信息内核,平台,下载源以及目录地位:

➜  ~ conda info  
  
     active environment : None  
       user config file : /Users/liuyue/.condarc  
 populated config files : /Users/liuyue/miniforge3/.condarc  
          conda version : 4.9.2  
    conda-build version : not installed  
         python version : 3.9.2.final.0  
       virtual packages : __osx=11.2.2=0  
                          __unix=0=0  
                          __archspec=1=arm64  
       base environment : /Users/liuyue/miniforge3  (read only)  
           channel URLs : https://conda.anaconda.org/conda-forge/osx-arm64  
                          https://conda.anaconda.org/conda-forge/noarch  
          package cache : /Users/liuyue/miniforge3/pkgs  
                          /Users/liuyue/.conda/pkgs  
       envs directories : /Users/liuyue/.conda/envs  
                          /Users/liuyue/miniforge3/envs  
               platform : osx-arm64  
             user-agent : conda/4.9.2 requests/2.25.1 CPython/3.9.2 Darwin/20.3.0 OSX/11.2.2  
                UID:GID : 502:20  
             netrc file : None  
           offline mode : False

因为一些家喻户晓的学术问题,咱们须要配置一下国内下载源:

conda config --add channels https://mirrors.ustc.edu.cn/anaconda/pkgs/main/  
conda config --add channels https://mirrors.ustc.edu.cn/anaconda/pkgs/free/  
conda config --add channels https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge/  
conda config --add channels https://mirrors.ustc.edu.cn/anaconda/cloud/msys2/  
conda config --add channels https://mirrors.ustc.edu.cn/anaconda/cloud/bioconda/  
conda config --add channels https://mirrors.ustc.edu.cn/anaconda/cloud/menpo/  
  
conda config --set show_channel_urls yes

随后查看以后下载源:

conda config --show

能够看到国内源曾经被增加进去了:

channel_priority: flexible  
channels:  
  - https://mirrors.ustc.edu.cn/anaconda/cloud/menpo/  
  - https://mirrors.ustc.edu.cn/anaconda/cloud/bioconda/  
  - https://mirrors.ustc.edu.cn/anaconda/cloud/msys2/  
  - https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge/  
  - https://mirrors.ustc.edu.cn/anaconda/pkgs/free/  
  - https://mirrors.ustc.edu.cn/anaconda/pkgs/main/  
  - defaults  
  - conda-forge  
client_ssl_cert: None

其余的一些 conda 常用命令:

1. conda –version #查看 conda 版本,验证是否装置

2. conda update conda #更新至最新版本,也会更新其它相干包

3. conda update –all #更新所有包

4. conda update package\_name #更新指定的包

5. conda create -n env\_name package\_name #创立名为 env\_name 的新环境,并在该环境下装置名为 package\_name 的包,能够指定新环境的版本号,例如:conda create -n python2 python=python2.7 numpy pandas,创立了 python2 环境,python 版本为 2.7,同时还装置了 numpy pandas 包

6. source activate env\_name #切换至 env\_name 环境

7. source deactivate #退出环境

8. conda info -e #显示所有曾经创立的环境

9. conda create –name new\_env\_name –clone old\_env\_name #复制 old\_env\_name 为 new\_env\_name

10. conda remove –name env\_name –all #删除环境

11. conda list #查看所有曾经装置的包

12. conda install package\_name #在以后环境中安装包

13. conda install –name env\_name package\_name #在指定环境中安装包

14. conda remove — name env\_name package #删除指定环境中的包

15. conda remove package #删除以后环境中的包

16. conda create -n tensorflow\_env tensorflow

conda activate tensorflow\_env #conda 装置 tensorflow 的 CPU 版本

17. conda create -n tensorflow\_gpuenv tensorflow-gpu

conda activate tensorflow\_gpuenv #conda 装置 tensorflow 的 GPU 版本

18. conda env remove -n env\_name #采纳第 10 条的办法删除环境失败时,可采纳这种办法

接着咱们来尝试集成深度学习框架 Tensorflow,因为目前默认是 3.9,咱们应用 conda 创立一个 3.8 的虚构开发环境:

sudo conda create -n py38 python=3.8

装置胜利后,输出命令:

conda info -e

就能够查看以后 conda 装置的所有版本:

➜  ~ conda info -e  
# conda environments:  
#  
base                  *  /Users/liuyue/miniforge3  
py38                     /Users/liuyue/miniforge3/envs/py38

能够看到一个默认的 3.9 环境,和新装的 3.8 环境,星号代表以后所处的环境,这里咱们切换到 3.8:

conda activate py38

此时环境曾经切换到 3.8:

(py38) ➜  ~ conda activate py38   
(py38) ➜  ~ conda info -e  
# conda environments:  
#  
base                     /Users/liuyue/miniforge3  
py38                  *  /Users/liuyue/miniforge3/envs/py38  
  
(py38) ➜  ~

上面开启深度学习框架 Tensorflow 之旅,因为苹果对 m1 芯片独自做了适配,所以不能用以前的 pip 形式间接进行装置,须要独自下载文件:https://github.91chifun.worke…://github.com/apple/tensorflow\_macos/releases/download/v0.1alpha1/tensorflow\_macos-0.1alpha1.tar.gz

解压文件:

tar -zxvf tensorflow_macos-0.1alpha1.tar.gz

解压后进入目录(肯定要进入 arm64 的文件内):

cd tensorflow_macos/arm64

执行命令利用下载的 arm64 内核安装包进行装置:

pip install --force pip==20.2.4 wheel setuptools cached-property six  
  
pip install --upgrade --no-dependencies --force numpy-1.18.5-cp38-cp38-macosx_11_0_arm64.whl grpcio-1.33.2-cp38-cp38-macosx_11_0_arm64.whl h5py-2.10.0-cp38-cp38-macosx_11_0_arm64.whl tensorflow_addons-0.11.2+mlcompute-cp38-cp38-macosx_11_0_arm64.whl  
  
pip install absl-py astunparse flatbuffers gast google_pasta keras_preprocessing opt_einsum protobuf tensorflow_estimator termcolor typing_extensions wrapt wheel tensorboard typeguard  
  
pip install --upgrade --force --no-dependencies tensorflow_macos-0.1a1-cp38-cp38-macosx_11_0_arm64.whl

装置胜利后,测试一下:

(py38) ➜  arm64 python  
Python 3.8.8 | packaged by conda-forge | (default, Feb 20 2021, 15:50:57)   
[Clang 11.0.1] on darwin  
Type "help", "copyright", "credits" or "license" for more information.  
>>> import tensorflow  
>>>

没有任何问题。

上面咱们来测试一下 M1 通过 Tensorflow 训练模型的效率,还记得衣香鬓影的“机械姬”吗:人工智能不过尔尔,基于 Python3 深度学习库 Keras/TensorFlow 打造属于本人的聊天机器人(ChatRobot)

编写 my\_chat.py:

import nltk  
import ssl  
from nltk.stem.lancaster import LancasterStemmer  
stemmer = LancasterStemmer()  
  
import numpy as np  
from tensorflow.keras.models import Sequential  
from tensorflow.keras.layers import Dense, Activation, Dropout  
from tensorflow.keras.optimizers import SGD  
import pandas as pd  
import pickle  
import random  
  
  
words = []  
classes = []  
documents = []  
ignore_words = ['?']  
# loop through each sentence in our intents patterns  
  
  
intents = {"intents": [  
        {"tag": "打招呼",  
         "patterns": ["你好", "您好", "请问", "有人吗", "徒弟","不好意思","美女","帅哥","靓妹"],  
         "responses": ["您好", "又是您啊", "吃了么您内","您有事吗"],  
         "context": [""]  
        },  
        {"tag": "辞别",  
         "patterns": ["再见", "拜拜", "88", "回见", "回头见"],  
         "responses": ["再见", "一路顺风", "下次见", "拜拜了您内"],  
         "context": [""]  
        },  
   ]  
}  
  
for intent in intents['intents']:  
    for pattern in intent['patterns']:  
        # tokenize each word in the sentence  
        w = nltk.word_tokenize(pattern)  
        # add to our words list  
        words.extend(w)  
        # add to documents in our corpus  
        documents.append((w, intent['tag']))  
        # add to our classes list  
        if intent['tag'] not in classes:  
            classes.append(intent['tag'])  
# stem and lower each word and remove duplicates  
words = [stemmer.stem(w.lower()) for w in words if w not in ignore_words]  
words = sorted(list(set(words)))  
# sort classes  
classes = sorted(list(set(classes)))  
# documents = combination between patterns and intents  
# print (len(documents), "documents")  
# # classes = intents  
# print (len(classes), "语境", classes)  
# # words = all words, vocabulary  
# print (len(words), "词数", words)  
  
  
# create our training data  
training = []  
# create an empty array for our output  
output_empty = [0] * len(classes)  
# training set, bag of words for each sentence  
for doc in documents:  
    # initialize our bag of words  
    bag = []  
  
    pattern_words = doc[0]  
     
    pattern_words = [stemmer.stem(word.lower()) for word in pattern_words]  
  
    for w in words:  
        bag.append(1) if w in pattern_words else bag.append(0)  
      
   
    output_row = list(output_empty)  
    output_row[classes.index(doc[1])] = 1  
      
    training.append([bag, output_row])  
  
random.shuffle(training)  
training = np.array(training)  
  
train_x = list(training[:,0])  
train_y = list(training[:,1])  
  
  
model = Sequential()  
model.add(Dense(128, input_shape=(len(train_x[0]),), activation='relu'))  
model.add(Dropout(0.5))  
model.add(Dense(64, activation='relu'))  
model.add(Dropout(0.5))  
model.add(Dense(len(train_y[0]), activation='softmax'))  
  
  
sgd = SGD(lr=0.01, decay=1e-6, momentum=0.9, nesterov=True)  
model.compile(loss='categorical_crossentropy', optimizer=sgd, metrics=['accuracy'])  
  
  
model.fit(np.array(train_x), np.array(train_y), epochs=200, batch_size=5, verbose=1)  
  
  
  
def clean_up_sentence(sentence):  
    # tokenize the pattern - split words into array  
    sentence_words = nltk.word_tokenize(sentence)  
    # stem each word - create short form for word  
    sentence_words = [stemmer.stem(word.lower()) for word in sentence_words]  
    return sentence_words  
# return bag of words array: 0 or 1 for each word in the bag that exists in the sentence  
def bow(sentence, words, show_details=True):  
    # tokenize the pattern  
    sentence_words = clean_up_sentence(sentence)  
    # bag of words - matrix of N words, vocabulary matrix  
    bag = [0]*len(words)    
    for s in sentence_words:  
        for i,w in enumerate(words):  
            if w == s:   
                # assign 1 if current word is in the vocabulary position  
                bag[i] = 1  
                if show_details:  
                    print ("found in bag: %s" % w)  
    return(np.array(bag))  
  
  
def classify_local(sentence):  
    ERROR_THRESHOLD = 0.25  
      
    # generate probabilities from the model  
    input_data = pd.DataFrame([bow(sentence, words)], dtype=float, index=['input'])  
    results = model.predict([input_data])[0]  
    # filter out predictions below a threshold, and provide intent index  
    results = [[i,r] for i,r in enumerate(results) if r>ERROR_THRESHOLD]  
    # sort by strength of probability  
    results.sort(key=lambda x: x[1], reverse=True)  
    return_list = []  
    for r in results:  
        return_list.append((classes[r[0]], str(r[1])))  
    # return tuple of intent and probability  
      
    return return_list  
  
  
p = bow("你好", words)  
print (p)  
  
print(classify_local('请问'))

返回后果:

(py38) ➜  mytornado git:(master) ✗ python3 test_mychat.py  
2021-03-03 22:43:21.059383: I tensorflow/compiler/mlir/mlir_graph_optimization_pass.cc:116] None of the MLIR optimization passes are enabled (registered 2)  
2021-03-03 22:43:21.059529: W tensorflow/core/platform/profile_utils/cpu_utils.cc:126] Failed to get CPU frequency: 0 Hz  
Epoch 1/200  
3/3 [==============================] - 0s 570us/step - loss: 0.6966 - accuracy: 0.5750  
Epoch 2/200  
3/3 [==============================] - 0s 482us/step - loss: 0.6913 - accuracy: 0.4857  
Epoch 3/200  
3/3 [==============================] - 0s 454us/step - loss: 0.6795 - accuracy: 0.4750  
Epoch 4/200  
3/3 [==============================] - 0s 434us/step - loss: 0.6913 - accuracy: 0.4750  
Epoch 5/200  
3/3 [==============================] - 0s 417us/step - loss: 0.6563 - accuracy: 0.5107  
Epoch 6/200  
3/3 [==============================] - 0s 454us/step - loss: 0.6775 - accuracy: 0.5714  
Epoch 7/200  
3/3 [==============================] - 0s 418us/step - loss: 0.6582 - accuracy: 0.6964  
Epoch 8/200  
3/3 [==============================] - 0s 487us/step - loss: 0.6418 - accuracy: 0.8071  
Epoch 9/200  
3/3 [==============================] - 0s 504us/step - loss: 0.6055 - accuracy: 0.6964  
Epoch 10/200  
3/3 [==============================] - 0s 457us/step - loss: 0.5933 - accuracy: 0.6964  
Epoch 11/200  
3/3 [==============================] - 0s 392us/step - loss: 0.6679 - accuracy: 0.5714  
Epoch 12/200  
3/3 [==============================] - 0s 427us/step - loss: 0.6060 - accuracy: 0.7464  
Epoch 13/200  
3/3 [==============================] - 0s 425us/step - loss: 0.6677 - accuracy: 0.5964  
Epoch 14/200  
3/3 [==============================] - 0s 420us/step - loss: 0.6208 - accuracy: 0.6214  
Epoch 15/200  
3/3 [==============================] - 0s 401us/step - loss: 0.6315 - accuracy: 0.6714  
Epoch 16/200  
3/3 [==============================] - 0s 401us/step - loss: 0.6718 - accuracy: 0.6464  
Epoch 17/200  
3/3 [==============================] - 0s 386us/step - loss: 0.6407 - accuracy: 0.6714  
Epoch 18/200  
3/3 [==============================] - 0s 505us/step - loss: 0.6031 - accuracy: 0.6464  
Epoch 19/200  
3/3 [==============================] - 0s 407us/step - loss: 0.6245 - accuracy: 0.6214  
Epoch 20/200  
3/3 [==============================] - 0s 422us/step - loss: 0.5805 - accuracy: 0.6964  
Epoch 21/200  
3/3 [==============================] - 0s 379us/step - loss: 0.6923 - accuracy: 0.5464  
Epoch 22/200  
3/3 [==============================] - 0s 396us/step - loss: 0.6383 - accuracy: 0.5714  
Epoch 23/200  
3/3 [==============================] - 0s 427us/step - loss: 0.6628 - accuracy: 0.5714  
Epoch 24/200  
3/3 [==============================] - 0s 579us/step - loss: 0.6361 - accuracy: 0.5964  
Epoch 25/200  
3/3 [==============================] - 0s 378us/step - loss: 0.5632 - accuracy: 0.7214  
Epoch 26/200  
3/3 [==============================] - 0s 387us/step - loss: 0.6851 - accuracy: 0.5214  
Epoch 27/200  
3/3 [==============================] - 0s 393us/step - loss: 0.6012 - accuracy: 0.6214  
Epoch 28/200  
3/3 [==============================] - 0s 392us/step - loss: 0.6470 - accuracy: 0.5964  
Epoch 29/200  
3/3 [==============================] - 0s 348us/step - loss: 0.6346 - accuracy: 0.6214  
Epoch 30/200  
3/3 [==============================] - 0s 362us/step - loss: 0.6350 - accuracy: 0.4964  
Epoch 31/200  
3/3 [==============================] - 0s 369us/step - loss: 0.5842 - accuracy: 0.5964  
Epoch 32/200  
3/3 [==============================] - 0s 481us/step - loss: 0.5279 - accuracy: 0.7214  
Epoch 33/200  
3/3 [==============================] - 0s 439us/step - loss: 0.5956 - accuracy: 0.7321  
Epoch 34/200  
3/3 [==============================] - 0s 355us/step - loss: 0.5570 - accuracy: 0.6964  
Epoch 35/200  
3/3 [==============================] - 0s 385us/step - loss: 0.5546 - accuracy: 0.8071  
Epoch 36/200  
3/3 [==============================] - 0s 375us/step - loss: 0.5616 - accuracy: 0.6714  
Epoch 37/200  
3/3 [==============================] - 0s 379us/step - loss: 0.6955 - accuracy: 0.6464  
Epoch 38/200  
3/3 [==============================] - 0s 389us/step - loss: 0.6089 - accuracy: 0.7321  
Epoch 39/200  
3/3 [==============================] - 0s 375us/step - loss: 0.5377 - accuracy: 0.6714  
Epoch 40/200  
3/3 [==============================] - 0s 392us/step - loss: 0.6224 - accuracy: 0.7179  
Epoch 41/200  
3/3 [==============================] - 0s 379us/step - loss: 0.6234 - accuracy: 0.5464  
Epoch 42/200  
3/3 [==============================] - 0s 411us/step - loss: 0.5224 - accuracy: 0.8321  
Epoch 43/200  
3/3 [==============================] - 0s 386us/step - loss: 0.5848 - accuracy: 0.5964  
Epoch 44/200  
3/3 [==============================] - 0s 401us/step - loss: 0.4620 - accuracy: 0.8679  
Epoch 45/200  
3/3 [==============================] - 0s 365us/step - loss: 0.4664 - accuracy: 0.8071  
Epoch 46/200  
3/3 [==============================] - 0s 367us/step - loss: 0.5904 - accuracy: 0.7679  
Epoch 47/200  
3/3 [==============================] - 0s 359us/step - loss: 0.5111 - accuracy: 0.7929  
Epoch 48/200  
3/3 [==============================] - 0s 363us/step - loss: 0.4712 - accuracy: 0.8679  
Epoch 49/200  
3/3 [==============================] - 0s 401us/step - loss: 0.5601 - accuracy: 0.8071  
Epoch 50/200  
3/3 [==============================] - 0s 429us/step - loss: 0.4884 - accuracy: 0.7929  
Epoch 51/200  
3/3 [==============================] - 0s 377us/step - loss: 0.5137 - accuracy: 0.8286  
Epoch 52/200  
3/3 [==============================] - 0s 368us/step - loss: 0.5475 - accuracy: 0.8286  
Epoch 53/200  
3/3 [==============================] - 0s 592us/step - loss: 0.4077 - accuracy: 0.8536  
Epoch 54/200  
3/3 [==============================] - 0s 400us/step - loss: 0.5367 - accuracy: 0.8179  
Epoch 55/200  
3/3 [==============================] - 0s 399us/step - loss: 0.5288 - accuracy: 0.8429  
Epoch 56/200  
3/3 [==============================] - 0s 367us/step - loss: 0.5775 - accuracy: 0.6964  
Epoch 57/200  
3/3 [==============================] - 0s 372us/step - loss: 0.5680 - accuracy: 0.6821  
Epoch 58/200  
3/3 [==============================] - 0s 360us/step - loss: 0.5164 - accuracy: 0.7321  
Epoch 59/200  
3/3 [==============================] - 0s 364us/step - loss: 0.5334 - accuracy: 0.6571  
Epoch 60/200  
3/3 [==============================] - 0s 358us/step - loss: 0.3858 - accuracy: 0.9036  
Epoch 61/200  
3/3 [==============================] - 0s 356us/step - loss: 0.4313 - accuracy: 0.8679  
Epoch 62/200  
3/3 [==============================] - 0s 373us/step - loss: 0.5017 - accuracy: 0.8429  
Epoch 63/200  
3/3 [==============================] - 0s 346us/step - loss: 0.4649 - accuracy: 0.8429  
Epoch 64/200  
3/3 [==============================] - 0s 397us/step - loss: 0.3804 - accuracy: 0.8893  
Epoch 65/200  
3/3 [==============================] - 0s 361us/step - loss: 0.5030 - accuracy: 0.7929  
Epoch 66/200  
3/3 [==============================] - 0s 372us/step - loss: 0.3958 - accuracy: 0.9286  
Epoch 67/200  
3/3 [==============================] - 0s 345us/step - loss: 0.4240 - accuracy: 0.8536  
Epoch 68/200  
3/3 [==============================] - 0s 360us/step - loss: 0.4651 - accuracy: 0.7929  
Epoch 69/200  
3/3 [==============================] - 0s 376us/step - loss: 0.4687 - accuracy: 0.7571  
Epoch 70/200  
3/3 [==============================] - 0s 398us/step - loss: 0.4660 - accuracy: 0.8429  
Epoch 71/200  
3/3 [==============================] - 0s 368us/step - loss: 0.3960 - accuracy: 0.9393  
Epoch 72/200  
3/3 [==============================] - 0s 355us/step - loss: 0.5523 - accuracy: 0.6071  
Epoch 73/200  
3/3 [==============================] - 0s 361us/step - loss: 0.5266 - accuracy: 0.7821  
Epoch 74/200  
3/3 [==============================] - 0s 371us/step - loss: 0.4245 - accuracy: 0.9643  
Epoch 75/200  
3/3 [==============================] - 0s 367us/step - loss: 0.5024 - accuracy: 0.7786  
Epoch 76/200  
3/3 [==============================] - 0s 453us/step - loss: 0.3419 - accuracy: 0.9393  
Epoch 77/200  
3/3 [==============================] - 0s 405us/step - loss: 0.4930 - accuracy: 0.7429  
Epoch 78/200  
3/3 [==============================] - 0s 672us/step - loss: 0.3443 - accuracy: 0.9036  
Epoch 79/200  
3/3 [==============================] - 0s 386us/step - loss: 0.3864 - accuracy: 0.8893  
Epoch 80/200  
3/3 [==============================] - 0s 386us/step - loss: 0.3863 - accuracy: 0.9286  
Epoch 81/200  
3/3 [==============================] - 0s 391us/step - loss: 0.2771 - accuracy: 0.8679  
Epoch 82/200  
3/3 [==============================] - 0s 370us/step - loss: 0.6083 - accuracy: 0.5571  
Epoch 83/200  
3/3 [==============================] - 0s 387us/step - loss: 0.2801 - accuracy: 0.9393  
Epoch 84/200  
3/3 [==============================] - 0s 357us/step - loss: 0.2483 - accuracy: 0.9286  
Epoch 85/200  
3/3 [==============================] - 0s 355us/step - loss: 0.2511 - accuracy: 0.9643  
Epoch 86/200  
3/3 [==============================] - 0s 339us/step - loss: 0.3410 - accuracy: 0.8893  
Epoch 87/200  
3/3 [==============================] - 0s 361us/step - loss: 0.3432 - accuracy: 0.9036  
Epoch 88/200  
3/3 [==============================] - 0s 347us/step - loss: 0.3819 - accuracy: 0.8893  
Epoch 89/200  
3/3 [==============================] - 0s 361us/step - loss: 0.5142 - accuracy: 0.7179  
Epoch 90/200  
3/3 [==============================] - 0s 502us/step - loss: 0.3055 - accuracy: 0.9393  
Epoch 91/200  
3/3 [==============================] - 0s 377us/step - loss: 0.3144 - accuracy: 0.8536  
Epoch 92/200  
3/3 [==============================] - 0s 376us/step - loss: 0.3712 - accuracy: 0.9036  
Epoch 93/200  
3/3 [==============================] - 0s 389us/step - loss: 0.1974 - accuracy: 0.9393  
Epoch 94/200  
3/3 [==============================] - 0s 365us/step - loss: 0.3128 - accuracy: 0.9393  
Epoch 95/200  
3/3 [==============================] - 0s 376us/step - loss: 0.2194 - accuracy: 1.0000  
Epoch 96/200  
3/3 [==============================] - 0s 377us/step - loss: 0.1994 - accuracy: 1.0000  
Epoch 97/200  
3/3 [==============================] - 0s 360us/step - loss: 0.1734 - accuracy: 0.9643  
Epoch 98/200  
3/3 [==============================] - 0s 367us/step - loss: 0.1786 - accuracy: 1.0000  
Epoch 99/200  
3/3 [==============================] - 0s 358us/step - loss: 0.4158 - accuracy: 0.8286  
Epoch 100/200  
3/3 [==============================] - 0s 354us/step - loss: 0.3131 - accuracy: 0.7571  
Epoch 101/200  
3/3 [==============================] - 0s 350us/step - loss: 0.1953 - accuracy: 0.8893  
Epoch 102/200  
3/3 [==============================] - 0s 403us/step - loss: 0.2577 - accuracy: 0.8429  
Epoch 103/200  
3/3 [==============================] - 0s 417us/step - loss: 0.2648 - accuracy: 0.8893  
Epoch 104/200  
3/3 [==============================] - 0s 377us/step - loss: 0.2901 - accuracy: 0.8286  
Epoch 105/200  
3/3 [==============================] - 0s 383us/step - loss: 0.2822 - accuracy: 0.9393  
Epoch 106/200  
3/3 [==============================] - 0s 381us/step - loss: 0.2837 - accuracy: 0.9036  
Epoch 107/200  
3/3 [==============================] - 0s 382us/step - loss: 0.3064 - accuracy: 0.8536  
Epoch 108/200  
3/3 [==============================] - 0s 352us/step - loss: 0.3376 - accuracy: 0.9036  
Epoch 109/200  
3/3 [==============================] - 0s 376us/step - loss: 0.3412 - accuracy: 0.8536  
Epoch 110/200  
3/3 [==============================] - 0s 363us/step - loss: 0.1718 - accuracy: 1.0000  
Epoch 111/200  
3/3 [==============================] - 0s 347us/step - loss: 0.1899 - accuracy: 0.8786  
Epoch 112/200  
3/3 [==============================] - 0s 363us/step - loss: 0.2352 - accuracy: 0.8286  
Epoch 113/200  
3/3 [==============================] - 0s 373us/step - loss: 0.1378 - accuracy: 1.0000  
Epoch 114/200  
3/3 [==============================] - 0s 353us/step - loss: 0.4288 - accuracy: 0.7071  
Epoch 115/200  
3/3 [==============================] - 0s 456us/step - loss: 0.4202 - accuracy: 0.6821  
Epoch 116/200  
3/3 [==============================] - 0s 382us/step - loss: 0.2962 - accuracy: 0.8893  
Epoch 117/200  
3/3 [==============================] - 0s 394us/step - loss: 0.2571 - accuracy: 0.8893  
Epoch 118/200  
3/3 [==============================] - 0s 365us/step - loss: 0.2697 - accuracy: 1.0000  
Epoch 119/200  
3/3 [==============================] - 0s 358us/step - loss: 0.3102 - accuracy: 0.9036  
Epoch 120/200  
3/3 [==============================] - 0s 367us/step - loss: 0.2928 - accuracy: 0.8286  
Epoch 121/200  
3/3 [==============================] - 0s 374us/step - loss: 0.3157 - accuracy: 0.8286  
Epoch 122/200  
3/3 [==============================] - 0s 381us/step - loss: 0.3920 - accuracy: 0.7786  
Epoch 123/200  
3/3 [==============================] - 0s 335us/step - loss: 0.2090 - accuracy: 0.9036  
Epoch 124/200  
3/3 [==============================] - 0s 368us/step - loss: 0.5079 - accuracy: 0.7786  
Epoch 125/200  
3/3 [==============================] - 0s 337us/step - loss: 0.1900 - accuracy: 0.9393  
Epoch 126/200  
3/3 [==============================] - 0s 339us/step - loss: 0.2047 - accuracy: 0.9643  
Epoch 127/200  
3/3 [==============================] - 0s 479us/step - loss: 0.3705 - accuracy: 0.7679  
Epoch 128/200  
3/3 [==============================] - 0s 390us/step - loss: 0.1850 - accuracy: 0.9036  
Epoch 129/200  
3/3 [==============================] - 0s 642us/step - loss: 0.1594 - accuracy: 0.9393  
Epoch 130/200  
3/3 [==============================] - 0s 373us/step - loss: 0.2010 - accuracy: 0.8893  
Epoch 131/200  
3/3 [==============================] - 0s 369us/step - loss: 0.0849 - accuracy: 1.0000  
Epoch 132/200  
3/3 [==============================] - 0s 349us/step - loss: 0.1145 - accuracy: 1.0000  
Epoch 133/200  
3/3 [==============================] - 0s 360us/step - loss: 0.1796 - accuracy: 1.0000  
Epoch 134/200  
3/3 [==============================] - 0s 371us/step - loss: 0.2363 - accuracy: 0.8536  
Epoch 135/200  
3/3 [==============================] - 0s 386us/step - loss: 0.1922 - accuracy: 0.9393  
Epoch 136/200  
3/3 [==============================] - 0s 369us/step - loss: 0.3595 - accuracy: 0.7679  
Epoch 137/200  
3/3 [==============================] - 0s 369us/step - loss: 0.1506 - accuracy: 0.8893  
Epoch 138/200  
3/3 [==============================] - 0s 377us/step - loss: 0.2471 - accuracy: 0.8536  
Epoch 139/200  
3/3 [==============================] - 0s 417us/step - loss: 0.1768 - accuracy: 0.8536  
Epoch 140/200  
3/3 [==============================] - 0s 400us/step - loss: 0.2112 - accuracy: 0.9393  
Epoch 141/200  
3/3 [==============================] - 0s 377us/step - loss: 0.3652 - accuracy: 0.7179  
Epoch 142/200  
3/3 [==============================] - 0s 364us/step - loss: 0.3007 - accuracy: 0.8429  
Epoch 143/200  
3/3 [==============================] - 0s 361us/step - loss: 0.0518 - accuracy: 1.0000  
Epoch 144/200  
3/3 [==============================] - 0s 373us/step - loss: 0.2144 - accuracy: 0.8286  
Epoch 145/200  
3/3 [==============================] - 0s 353us/step - loss: 0.0888 - accuracy: 1.0000  
Epoch 146/200  
3/3 [==============================] - 0s 361us/step - loss: 0.1267 - accuracy: 1.0000  
Epoch 147/200  
3/3 [==============================] - 0s 341us/step - loss: 0.0321 - accuracy: 1.0000  
Epoch 148/200  
3/3 [==============================] - 0s 358us/step - loss: 0.0860 - accuracy: 1.0000  
Epoch 149/200  
3/3 [==============================] - 0s 375us/step - loss: 0.2151 - accuracy: 0.8893  
Epoch 150/200  
3/3 [==============================] - 0s 351us/step - loss: 0.1592 - accuracy: 1.0000  
Epoch 151/200  
3/3 [==============================] - 0s 531us/step - loss: 0.1450 - accuracy: 0.8786  
Epoch 152/200  
3/3 [==============================] - 0s 392us/step - loss: 0.1813 - accuracy: 0.9036  
Epoch 153/200  
3/3 [==============================] - 0s 404us/step - loss: 0.1197 - accuracy: 1.0000  
Epoch 154/200  
3/3 [==============================] - 0s 367us/step - loss: 0.0930 - accuracy: 1.0000  
Epoch 155/200  
3/3 [==============================] - 0s 580us/step - loss: 0.2587 - accuracy: 0.8893  
Epoch 156/200  
3/3 [==============================] - 0s 383us/step - loss: 0.0742 - accuracy: 1.0000  
Epoch 157/200  
3/3 [==============================] - 0s 353us/step - loss: 0.1197 - accuracy: 0.9643  
Epoch 158/200  
3/3 [==============================] - 0s 371us/step - loss: 0.1716 - accuracy: 0.8536  
Epoch 159/200  
3/3 [==============================] - 0s 337us/step - loss: 0.1300 - accuracy: 0.9643  
Epoch 160/200  
3/3 [==============================] - 0s 347us/step - loss: 0.1439 - accuracy: 0.9393  
Epoch 161/200  
3/3 [==============================] - 0s 366us/step - loss: 0.2597 - accuracy: 0.9393  
Epoch 162/200  
3/3 [==============================] - 0s 345us/step - loss: 0.1605 - accuracy: 0.8893  
Epoch 163/200  
3/3 [==============================] - 0s 468us/step - loss: 0.0437 - accuracy: 1.0000  
Epoch 164/200  
3/3 [==============================] - 0s 372us/step - loss: 0.0376 - accuracy: 1.0000  
Epoch 165/200  
3/3 [==============================] - 0s 391us/step - loss: 0.0474 - accuracy: 1.0000  
Epoch 166/200  
3/3 [==============================] - 0s 378us/step - loss: 0.3225 - accuracy: 0.7786  
Epoch 167/200  
3/3 [==============================] - 0s 368us/step - loss: 0.0770 - accuracy: 1.0000  
Epoch 168/200  
3/3 [==============================] - 0s 367us/step - loss: 0.5629 - accuracy: 0.7786  
Epoch 169/200  
3/3 [==============================] - 0s 359us/step - loss: 0.0177 - accuracy: 1.0000  
Epoch 170/200  
3/3 [==============================] - 0s 370us/step - loss: 0.1167 - accuracy: 1.0000  
Epoch 171/200  
3/3 [==============================] - 0s 349us/step - loss: 0.1313 - accuracy: 1.0000  
Epoch 172/200  
3/3 [==============================] - 0s 337us/step - loss: 0.0852 - accuracy: 0.9393  
Epoch 173/200  
3/3 [==============================] - 0s 375us/step - loss: 0.0545 - accuracy: 1.0000  
Epoch 174/200  
3/3 [==============================] - 0s 354us/step - loss: 0.0674 - accuracy: 0.9643  
Epoch 175/200  
3/3 [==============================] - 0s 355us/step - loss: 0.0911 - accuracy: 1.0000  
Epoch 176/200  
3/3 [==============================] - 0s 404us/step - loss: 0.0980 - accuracy: 0.9393  
Epoch 177/200  
3/3 [==============================] - 0s 396us/step - loss: 0.0465 - accuracy: 1.0000  
Epoch 178/200  
3/3 [==============================] - 0s 403us/step - loss: 0.1117 - accuracy: 0.9393  
Epoch 179/200  
3/3 [==============================] - 0s 373us/step - loss: 0.0415 - accuracy: 1.0000  
Epoch 180/200  
3/3 [==============================] - 0s 369us/step - loss: 0.0825 - accuracy: 1.0000  
Epoch 181/200  
3/3 [==============================] - 0s 425us/step - loss: 0.0378 - accuracy: 1.0000  
Epoch 182/200  
3/3 [==============================] - 0s 381us/step - loss: 0.1155 - accuracy: 0.9393  
Epoch 183/200  
3/3 [==============================] - 0s 354us/step - loss: 0.0207 - accuracy: 1.0000  
Epoch 184/200  
3/3 [==============================] - 0s 346us/step - loss: 0.0344 - accuracy: 1.0000  
Epoch 185/200  
3/3 [==============================] - 0s 379us/step - loss: 0.0984 - accuracy: 0.9393  
Epoch 186/200  
3/3 [==============================] - 0s 360us/step - loss: 0.1508 - accuracy: 0.8536  
Epoch 187/200  
3/3 [==============================] - 0s 361us/step - loss: 0.0463 - accuracy: 1.0000  
Epoch 188/200  
3/3 [==============================] - 0s 358us/step - loss: 0.0476 - accuracy: 0.9643  
Epoch 189/200  
3/3 [==============================] - 0s 379us/step - loss: 0.1592 - accuracy: 1.0000  
Epoch 190/200  
3/3 [==============================] - 0s 387us/step - loss: 0.0071 - accuracy: 1.0000  
Epoch 191/200  
3/3 [==============================] - 0s 405us/step - loss: 0.0527 - accuracy: 1.0000  
Epoch 192/200  
3/3 [==============================] - 0s 401us/step - loss: 0.0874 - accuracy: 0.9393  
Epoch 193/200  
3/3 [==============================] - 0s 355us/step - loss: 0.0199 - accuracy: 1.0000  
Epoch 194/200  
3/3 [==============================] - 0s 373us/step - loss: 0.1299 - accuracy: 0.9643  
Epoch 195/200  
3/3 [==============================] - 0s 360us/step - loss: 0.0929 - accuracy: 1.0000  
Epoch 196/200  
3/3 [==============================] - 0s 380us/step - loss: 0.0265 - accuracy: 1.0000  
Epoch 197/200  
3/3 [==============================] - 0s 358us/step - loss: 0.0843 - accuracy: 1.0000  
Epoch 198/200  
3/3 [==============================] - 0s 354us/step - loss: 0.0925 - accuracy: 1.0000  
Epoch 199/200  
3/3 [==============================] - 0s 327us/step - loss: 0.0770 - accuracy: 1.0000  
Epoch 200/200  
3/3 [==============================] - 0s 561us/step - loss: 0.0311 - accuracy: 1.0000  
found in bag: 你好  
[0 0 1 0 0 0 0 0 0 0 0 0 0 0]  
found in bag: 请问  
[('打招呼', '0.998965')]

霎时执行结束,秒杀 intel 芯片的 mac,怎一个香字了得!

接下来,尝试装置另外一个在业界名声煊赫的深度学习框架 Pytorch!

因为以后 arm64 架构的只反对 3.9 版本,所以咱们来创立一个虚拟空间:

sudo conda create -n pytorch numpy matplotlib pandas python=3.9

这里提前将须要的根底库都一一装置,因为如果不在创立虚拟空间时提前装置,之后应用 pip 是装置不上的,装置胜利后,激活环境:

(pytorch) ➜  conda activate pytorch                                          
(pytorch) ➜

随后下载 arm64 版本的 pytorch 安装包:https://github.com/wizyoung/A…\_11\_0\_arm64.whl

下载胜利后,执行装置命令:

sudo pip install torch-1.8.0a0-cp39-cp39-macosx_11_0_arm64.whl

让咱们来试试 Pytorch 在 M1 芯片加持后的性能,编写 test\_torch.py:

from tqdm import tqdm  
import torch  
  
@torch.jit.script  
def foo():  
    x = torch.ones((1024 * 12, 1024 * 12), dtype=torch.float32)  
    y = torch.ones((1024 * 12, 1024 * 12), dtype=torch.float32)  
    z = x + y  
    return z  
  
  
if __name__ == '__main__':  
    z0 = None  
    for _ in tqdm(range(10000000000)):  
        zz = foo()  
        if z0 is None:  
            z0 = zz  
        else:  
            z0 += zz

后果仍然秒杀 Intel mac,达到了 320 it/s,你敢信?

最初,有没有 arm64 架构的编辑器呢?答案是有的,vscode 值得领有,下载地址:https://code.visualstudio.com… 肯定要抉择 arm64 版的:

解压后间接运行即可,能够在插件商店抉择 Python 和 Code Runner,即可开启 M1 的 Python 代码编写之旅。

结语:M1 芯片的 Mac 和 Python3,几乎就是金风玉露,绝配天成。只有撩开 M1 和开发者们之间的那一层帷幔,期待咱们的,就是纵享丝滑的开发感触,还等什么?犹豫只会败北,是时候焚烧灵魂,献出钱包了。

原文转载自「刘悦的技术博客」https://v3u.cn/a_id_189

退出移动版