Fix #44: it now runs on M1 macOS using tensorflow-metal

README.md

@@ -86,11 +86,11 @@ You have two alternative package managers you can use to install `nlpia`:
 
 5.1. [`conda`](https://github.com/totalgood/nlpia/blob/master/README.md#alternative-51-conda)
 5.2. [`pip`](https://github.com/totalgood/nlpia/blob/master/README.md#alternative-52-pip)
+5.3. [`docker`](https://github.com/totalgood/nlpia/blob/master/README.md#alternative-53-docker)
+5.4. [`miniforge`](https://github.com/totalgood/nlpia/blob/master/README.md#alternative-54-miniforge)
 
 A helpful [NLPIA](http://bit.ly/gh-readme-nlpia-book) reader, [Hoang Chung Hien](https://github.com/hoangchunghien), created a Dockerfile you can use for a third way to manage your environment:
 
-5.3. [`docker`](https://github.com/totalgood/nlpia/blob/master/README.md#alternative-53-docker)
-
 In most cases, `conda` will be able to install python packages faster and more reliably than `pip`. Without `conda` Some packages, such as `python-levenshtein`, require you to compile a C library during installation. Windows doesn't have a a compiler and python package installer that will "just work."
 
 #### Alternative 5.1. `conda`
@@ -191,6 +191,25 @@ docker run -p 8888:8888 -v ~:/home/jovyan/work nlpia
 
 Then open a new notebook and test your code. Make sure save it inside `work` directory so it's accessible outside the container.
 
+
+## Alternative 5.4. `miniforge`
+
+### 5.4.1 Install Python3.9.5 and tensorflow-metal with miniforge
+
+Follow this [tutorial](https://makeoptim.com/en/deep-learning/tensorflow-metal)
+
+### 5.4.2 Install all requirements with pip
+
+Miniforge will give you a Python3.9.5 (or more) virtual environment, capable to run the accelerated tensorflow-metal , optimized to use the M1 integrated GPU hardware.
+
+    cd nlpia
+    pip install -r requirements.txt
+
+### 5.4.3 Install nlpia
+
+    python setup.py install
+
+
 ### 6. Have Fun!
 
 Check out the code examples from the book in `nlpia/nlpia/book/examples` to get ideas:

requirements/requirements.txt

@@ -2,7 +2,7 @@ AIML-Bot>=0.0.3,<0.1.0
 coverage>=4.3.4
 cython>=0.29.20
 future>=0.16.0
-gensim==3.6.0
+gensim
 html2text>=2018.1.9
 html5lib==1.1
 jupyter>=1.0.0,<1.1.0
@@ -12,8 +12,8 @@ Keras-Preprocessing>=1.0.5
 matplotlib>=3.2.2,<3.4.0
 MechanicalSoup>=0.10.0,<0.11.0
 nltk==3.5
-pandas>=0.25.3,<0.26.0
-pandoc==2.2.3.2
+pandas
+pandoc
 pathlib>=1.0.1,<2.0.0
 pillow==6.2.2
 pip>=18.0
@@ -27,12 +27,13 @@ qtconsole==4.4.1. # ,<6.0.0  # 4.4.1 is available on OSX in conda and pip
 redis>=2.10.6,<2.11.0
 regex>=2020.6.8
 scikit-learn>=0.18.1
-scipy==1.4.1
+scipy
 seaborn>=0.8
 setuptools>=40.8.0
-spacy==2.3.0
-tensorflow==2.1.0
+spacy
+tensorflow-metal
 tensorflow-hub>=0.4.0
 theano==1.0.2
 tqdm>=4.11.2
 wheel>=0.30.0a0
+ipdb

setup.cfg

@@ -54,7 +54,7 @@ install_requires =
     gensim
     pandas-datareader
     pugnlp
-    tensorflow
+    tensorflow-metal
     keras
     regex
     spacy
@@ -82,7 +82,7 @@ exclude =
 # Add here additional requirements for extra features, to install with:
 # `pip install nlpia[voice]` like:
 twitter = tweepy; twython
-deep = keras; tensorflow; annoy; gensim
+deep = keras; tensorflow-metal; annoy; gensim
 # voice = pyaudio; pocketsphinx; SpeechRecognition; pyttsx3; deepspeech
 chat = aichat; will; redis; ChatterBot; lxml; aiml; aichat
 all = nlpia[deep]; nlpia[chat]; nlpia[twitter]; nlpia[voice]