###python3.6.8###
###查询最新版本 https://pypi.org/search/?q=numpy
###或者$pip3 search tensorflow | egrep "^.*tensorflow.*\("
###usage(以下常用lib需要手工执行更新):
##让PIP源使用国内镜像，提升下载速度和安装成功率。https://www.cnblogs.com/microman/p/6107879.html
##别再使用pip安装TensorFlow了！用conda吧～ https://www.jiqizhixin.com/articles/2018-10-12-2
##pip或pi3安装$pip install -U tensorflow-gpu>=2.0.0-beta1 -i https://pypi.tuna.tsinghua.edu.cn/simple (清华加速pip源可用)
###手工升级$pip3 install -U tensorflow-gpu==2.0.0-beta1 -i https://pypi.douban.com/simple
###手工升级$pip3 install -U tensorflow-hub==0.1.1 -i https://pypi.douban.com/simple
###遵照https://pytorch.org/页面提示手工升级$pip3 install -U torch torchvision -i https://pypi.douban.com/simple
###手工升级$pip3 install -U numpy==1.15.2 -i https://pypi.douban.com/simple
###手工升级$pip3 install -U pandas==0.23.4 -i https://pypi.douban.com/simple
###手工升级(需要才升级):$sudo pip3 install -U h5py
##########################################################
# example: pip --no-cache-dir  install -U ray==0.8.4 -i https://pypi.tuna.tsinghua.edu.cn/simple
# for common, import requirements-common.txt
# for gpu, import requirements-gpu.txt
# for cpu, import requirements.txt
#################
# 2020-01-19日后生产环境需要更新的依赖为
# pandas==1.0.3 # 之前版本是0.25.2(joinquant不支持1.0.0以上的pandas版本)
# gym==0.15.4
# transformers==2.4.1
# ##tensorflow_hub==0.7.0 # 可以不用该库，用transformers代替
# ##sentencepiece==0.1.85 # 可以不用该库
# tensorflow-datasets==2.0.0
# cvxopt==1.2.4
# deap==1.3.1
# pygraphviz==1.5
# scikit_learn==0.22.2
# tpot==0.11.1
# xgboost==0.90 # mac os上用$brew install xgboost安装
#################