Alternative: use existing AMI
benchmarking:
http://www.bitfusion.io/2016/11/03/quick-comparison-of-tensorflow-gpu-performance-on-aws-p2-and-g2-instances/
http://www.bitfusion.io/2016/10/21/bitfusion-deep-learning-amis-now-available-on-aws-p2-instances/
http://www.bitfusion.io/2016/05/09/easy-tensorflow-model-training-aws/
https://github.com/bitfusionio/amis/tree/master/awsmrkt-bfboost-ubuntu14-cuda75-tensorflow
jcx9dy@donna:~$ ssh -i "donna.pem" [email protected]
The authenticity of host '54.235.54.22 (54.235.54.22)' can't be established.
ECDSA key fingerprint is SHA256:Duv7Jww2pjr7i4n3lpjBELQOq48hOk1+df9wh9TpBhA.
Are you sure you want to continue connecting (yes/no)? yes
Warning: Permanently added '54.235.54.22' (ECDSA) to the list of known hosts.
@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
@ WARNING: UNPROTECTED PRIVATE KEY FILE! @
@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
Permissions 0664 for 'donna.pem' are too open.
It is required that your private key files are NOT accessible by others.
This private key will be ignored.
Load key "donna.pem": bad permissions
Permission denied (publickey).
jcx9dy@donna:~$ chmod 400 donna.pem
jcx9dy@donna:~$ ssh -i "donna.pem" [email protected]
########################################################################################################################
########################################################################################################################
____ _ _ __ _ _
| __ )(_) |_ / _|_ _ ___(_) ___ _ __ (_) ___
| _ \| | __| |_| | | / __| |/ _ \| '_ \ | |/ _ \
| |_) | | |_| _| |_| \__ \ | (_) | | | |_| | (_) |
|____/|_|\__|_| \__,_|___/_|\___/|_| |_(_)_|\___/
Welcome to Bitfusion Ubuntu 14 Tensorflow - Ubuntu 14.04 LTS (GNU/Linux 3.13.0-101-generic x86_64)
This AMI is brought to you by Bitfusion.io
http://www.bitfusion.io
Please email all feedback and support requests to:
[email protected]
We would love to hear from you! Contact us with any feedback or a feature request at the email above.
########################################################################################################################
########################################################################################################################
Last login: Tue Nov 22 18:14:28 2016 from d-128-143-129-160.dhcp.virginia.edu
ubuntu@ip-172-31-60-241:~$ nvidia-smi
Tue Nov 22 18:15:05 2016
+------------------------------------------------------+
| NVIDIA-SMI 352.99 Driver Version: 352.99 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 Tesla K80 On | 0000:00:1E.0 Off | 0 |
| N/A 35C P8 30W / 149W | 55MiB / 11519MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
| No running processes found |
+-----------------------------------------------------------------------------+
watch -n0.1 nvidia-smi
ubuntu@ip-172-31-60-241:~$ top
top - 18:16:48 up 6 min, 1 user, load average: 0.13, 0.61, 0.36
Tasks: 141 total, 1 running, 140 sleeping, 0 stopped, 0 zombie
%Cpu0 : 0.0 us, 0.0 sy, 0.0 ni,100.0 id, 0.0 wa, 0.0 hi, 0.0 si, 0.0 st
%Cpu1 : 0.0 us, 0.0 sy, 0.0 ni,100.0 id, 0.0 wa, 0.0 hi, 0.0 si, 0.0 st
%Cpu2 : 0.0 us, 0.0 sy, 0.0 ni,100.0 id, 0.0 wa, 0.0 hi, 0.0 si, 0.0 st
%Cpu3 : 0.0 us, 0.0 sy, 0.0 ni,100.0 id, 0.0 wa, 0.0 hi, 0.0 si, 0.0 st
KiB Mem: 62881772 total, 918128 used, 61963644 free, 35080 buffers
KiB Swap: 0 total, 0 used, 0 free. 321184 cached Mem
ubuntu@ip-172-31-60-241:~$ df -h
Filesystem Size Used Avail Use% Mounted on
udev 30G 12K 30G 1% /dev
tmpfs 6.0G 796K 6.0G 1% /run
/dev/xvda1 99G 25G 70G 27% /
none 4.0K 0 4.0K 0% /sys/fs/cgroup
none 5.0M 0 5.0M 0% /run/lock
none 30G 12K 30G 1% /run/shm
none 100M 8.0K 100M 1% /run/user
ubuntu@ip-172-31-60-241:~$ ll
total 140
drwxr-xr-x 17 ubuntu ubuntu 4096 Nov 16 00:46 ./
drwxr-xr-x 4 root root 4096 Nov 15 21:41 ../
-rw-r--r-- 1 root root 0 Nov 16 00:46 .bash_history
-rw-r--r-- 1 ubuntu ubuntu 220 Apr 9 2014 .bash_logout
-rw-r--r-- 1 ubuntu ubuntu 3688 Nov 15 21:50 .bashrc
drwxr-xr-x 4 ubuntu ubuntu 4096 Nov 15 21:50 .bazel/
drwxr-xr-x 2 ubuntu ubuntu 4096 Nov 15 21:50 bin/
drwx------ 4 ubuntu ubuntu 4096 Nov 15 21:50 .cache/
drwxr-xr-x 3 ubuntu ubuntu 4096 Nov 15 21:41 .config/
-rw-rw-r-- 1 ubuntu ubuntu 38324 Nov 18 23:30 EULA
-rw------- 1 root root 0 Nov 16 00:46 .history
drwx------ 4 ubuntu ubuntu 4096 Nov 22 18:12 .jupyter/
drwxr-xr-x 2 ubuntu ubuntu 4096 Nov 16 00:45 .keras/
drwxr-xr-x 2 ubuntu ubuntu 4096 Nov 16 00:45 keras-examples/
drwxr-xr-x 3 ubuntu ubuntu 4096 Nov 15 21:41 .local/
drwxr-xr-x 6 ubuntu ubuntu 4096 Nov 15 23:58 magenta/
drwx------ 3 root root 4096 Nov 16 00:00 .nv/
drwxr-xr-x 2 ubuntu ubuntu 4096 Nov 15 21:50 .oracle_jre_usage/
-rw-r--r-- 1 ubuntu ubuntu 705 Nov 16 00:46 .profile
drwxr-xr-x 3 ubuntu ubuntu 4096 Nov 16 00:46 pynb/
-rw-rw-r-- 1 ubuntu ubuntu 12837 Nov 18 23:30 README
-rw------- 1 root root 1024 Nov 15 21:41 .rnd
drwx------ 2 ubuntu ubuntu 4096 Nov 15 00:03 .ssh/
drwxr-xr-x 7 ubuntu ubuntu 4096 Nov 15 22:55 tensorflow/
drwxr-xr-x 7 ubuntu ubuntu 4096 Nov 16 00:07 tf-serving/
ubuntu@ip-172-31-60-241:~$ uname -a
Linux ip-172-31-60-241 3.13.0-101-generic #148-Ubuntu SMP Thu Oct 20 22:08:32 UTC 2016 x86_64 x86_64 x86_64 GNU/Linux
ubuntu@ip-172-31-60-241:~$ which python
/usr/bin/python
ubuntu@ip-172-31-60-241:~$ which python3
/usr/bin/python3
ubuntu@ip-172-31-60-241:~$ python --version
Python 2.7.6
ubuntu@ip-172-31-60-241:~$ python3 --version
Python 3.4.3
python
Python 2.7.6 (default, Jun 22 2015, 17:58:13)
[GCC 4.8.2] on linux2
>>> import keras
Using TensorFlow backend.
I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcublas.so.7.5 locally
I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcudnn.so.5 locally
I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcufft.so.7.5 locally
I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcuda.so.1 locally
I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcurand.so.7.5 locally
>>> print keras.__version__
1.1.1
python3
Python 3.4.3 (default, Sep 14 2016, 12:36:27)
GCC 4.8.4] on linux
>>> import keras
Using TensorFlow backend.
I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcublas.so.7.5 locally
I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcudnn.so.5 locally
I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcufft.so.7.5 locally
I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcuda.so.1 locally
I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcurand.so.7.5 locally
>>> keras.__version__
'1.1.1'
ubuntu@ip-172-31-60-241:~$ python ~/tensorflow/tensorflow/models/image/cifar10/cifar10_multi_gpu_train.py
I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcublas.so.7.5 locally
I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcudnn.so.5 locally
I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcufft.so.7.5 locally
I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcuda.so.1 locally
I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcurand.so.7.5 locally
>> Downloading cifar-10-binary.tar.gz 100.0%
Successfully downloaded cifar-10-binary.tar.gz 170052171 bytes.
Filling queue with 20000 CIFAR images before starting to train. This will take a few minutes.
I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:925] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
I tensorflow/core/common_runtime/gpu/gpu_device.cc:951] Found device 0 with properties:
name: Tesla K80
major: 3 minor: 7 memoryClockRate (GHz) 0.8235
pciBusID 0000:00:1e.0
Total memory: 11.25GiB
Free memory: 11.13GiB
I tensorflow/core/common_runtime/gpu/gpu_device.cc:972] DMA: 0
I tensorflow/core/common_runtime/gpu/gpu_device.cc:982] 0: Y
I tensorflow/core/common_runtime/gpu/gpu_device.cc:1041] Creating TensorFlow device (/gpu:0) -> (device: 0, name: Tesla K80, pci bus id: 0000:00:1e.0)
2016-11-22 18:29:20.997334: step 0, loss = 4.67 (2.1 examples/sec; 62.301 sec/batch)
2016-11-22 18:29:23.064279: step 10, loss = 4.64 (840.6 examples/sec; 0.152 sec/batch)
2016-11-22 18:29:24.643143: step 20, loss = 4.51 (826.9 examples/sec; 0.155 sec/batch)
2016-11-22 18:29:26.254807: step 30, loss = 4.39 (770.8 examples/sec; 0.166 sec/batch)
2016-11-22 18:29:27.873522: step 40, loss = 4.40 (770.2 examples/sec; 0.166 sec/batch)
2016-11-22 18:29:29.533449: step 50, loss = 4.23 (766.2 examples/sec; 0.167 sec/batch)
2016-11-22 18:29:31.175779: step 60, loss = 4.33 (825.3 examples/sec; 0.155 sec/batch)
2016-11-22 18:29:32.838024: step 70, loss = 4.41 (788.5 examples/sec; 0.162 sec/batch)
2016-11-22 18:29:34.486316: step 80, loss = 4.40 (790.2 examples/sec; 0.162 sec/batch)
2016-11-22 18:29:36.158620: step 90, loss = 4.09 (744.8 examples/sec; 0.172 sec/batch)
2016-11-22 18:29:37.821699: step 100, loss = 4.12 (766.1 examples/sec; 0.167 sec/batch)
2016-11-22 18:29:54.621281: step 200, loss = 3.82 (783.4 examples/sec; 0.163 sec/batch)
2016-11-22 18:30:11.167037: step 300, loss = 3.56 (774.1 examples/sec; 0.165 sec/batch)
2016-11-22 18:30:27.910897: step 400, loss = 3.37 (763.4 examples/sec; 0.168 sec/batch)
2016-11-22 18:30:44.432482: step 500, loss = 3.09 (766.8 examples/sec; 0.167 sec/batch)
2016-11-22 18:32:07.944061: step 1000, loss = 2.53 (787.3 examples/sec; 0.163 sec/batch)
2016-11-22 18:34:56.223261: step 2000, loss = 1.50 (769.3 examples/sec; 0.166 sec/batch)
2016-11-22 18:37:43.819220: step 3000, loss = 1.48 (766.6 examples/sec; 0.167 sec/batch)
2016-11-22 18:40:30.868867: step 4000, loss = 1.15 (759.0 examples/sec; 0.169 sec/batch)
2016-11-22 18:43:17.959348: step 5000, loss = 0.79 (801.6 examples/sec; 0.160 sec/batch)
2016-11-22 18:46:03.535481: step 6000, loss = 0.98 (775.0 examples/sec; 0.165 sec/batch)
2016-11-22 18:48:50.086918: step 7000, loss = 0.94 (786.9 examples/sec; 0.163 sec/batch)
2016-11-22 18:51:35.410308: step 8000, loss = 1.26 (800.2 examples/sec; 0.160 sec/batch)
2016-11-22 18:54:20.427048: step 9000, loss = 0.95 (780.1 examples/sec; 0.164 sec/batch)
2016-11-22 18:57:06.321096: step 10000, loss = 0.96 (755.7 examples/sec; 0.169 sec/batch)
2016-11-22 18:59:52.777928: step 11000, loss = 0.83 (793.5 examples/sec; 0.161 sec/batch)
2016-11-22 19:02:38.845511: step 12000, loss = 0.86 (785.2 examples/sec; 0.163 sec/batch)
2016-11-22 19:05:24.809961: step 13000, loss = 0.95 (791.0 examples/sec; 0.162 sec/batch)
ubuntu@ip-172-31-60-241:~$ python ~/tensorflow/tensorflow/models/image/cifar10/cifar10_eval.py
2016-11-22 19:53:21.111682: precision @ 1 = 0.827
#copy model checkpoints to s3 (assumes aws config was set properly)
aws s3 cp /tmp s3://cs6501/cifar10/ --recursive
https://www.tensorflow.org/versions/r0.11/tutorials/deep_cnn/index.html
It consists of 1,068,298 learnable parameters and requires about 19.5M multiply-add operations to compute inference on a single image.
https://github.com/tensorflow/tensorflow/blob/master/tensorflow/models/image/cifar10/cifar10_multi_gpu_train.py
Accuracy:
cifar10_multi_gpu_train.py achieves ~86% accuracy after 100K steps (256
epochs of data) as judged by cifar10_eval.py.
Speed: With batch_size 128.
System | Step Time (sec/batch) | Accuracy
--------------------------------------------------------------------
1 Tesla K20m | 0.35-0.60 | ~86% at 60K steps (5 hours)
1 Tesla K40m | 0.25-0.35 | ~86% at 100K steps (4 hours)
2 Tesla K20m | 0.13-0.20 | ~84% at 30K steps (2.5 hours)
3 Tesla K20m | 0.13-0.18 | ~84% at 30K steps
4 Tesla K20m | ~0.10 | ~84% at 30K steps
New Group Security policy
TCP protocols:
enable port 8787
enable port 8888 for ipython
enable port 6006 for tensorboard
enable port 22 for ssh
enable port 80 for http
enable port 443 for https
Copy data to instance
aws s3 cp s3://cs6501/data/patches.tar.gz ~/data/
tar -xzf ~/data/patches.tar.gz &
Optimize AMI
sudo pip3 install git+git://github.com/fchollet/keras.git --upgrade
sudo nvidia-smi -pm 1
sudo nvidia-smi --auto-boost-default=0
sudo nvidia-smi -ac 2505,875
wget -O /usr/local/lib/python3.4/dist-packages/keras/preprocessing/image.py https://raw.githubusercontent.com/jfx319/vicente_CS6501/master/image.py?token=ATvGiJdgJoOzlSo16n_iT46IJCemHydaks5YTQeVwA%3D%3D
sudo kill $(pgrep jupyter)US East (N. Virginia) ami-bbd1e7accuda 8.0 , 16 GiB root volume that was sufficient for the installation http://expressionflow.com/2016/10/09/installing-tensorflow-on-an-aws-ec2-p2-gpu-instance/
https://github.com/saiprashanths/dl-docker#specs
http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/using_cluster_computing.html#install-nvidia-driver http://www.nvidia.com/object/gpu-accelerated-applications-tensorflow-installation.html
https://aws.amazon.com/blogs/aws/new-p2-instance-type-for-amazon-ec2-up-to-16-gpus/
http://ramhiser.com/2016/01/05/installing-tensorflow-on-an-aws-ec2-instance-with-gpu-support/
sudo apt-get update
sudo apt-get upgrade
sudo apt-get install -y build-essential git python-pip libfreetype6-dev libxft-dev libncurses-dev libopenblas-dev gfortran python-matplotlib libblas-dev liblapack-dev libatlas-base-dev python-dev python-pydot linux-headers-generic linux-image-extra-virtual unzip python-numpy swig python-pandas python-sklearn unzip wget pkg-config zip g++ zlib1g-dev
sudo pip install -U pipinstall CUDA Toolkit 7.0
wget http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1410/x86_64/cuda-repo-ubuntu1410_7.0-28_amd64.deb
sudo dpkg -i cuda-repo-ubuntu1410_7.0-28_amd64.deb
rm cuda-repo-ubuntu1410_7.0-28_amd64.deb
sudo apt-get update
sudo apt-get install -y cuda
#install CUDNN
#https://developer.nvidia.com/rdp/assets/cudnn-65-linux-v2-asset
tar -zxf cudnn-6.5-linux-x64-v2.tgz && rm cudnn-6.5-linux-x64-v2.tgz
sudo cp -R cudnn-6.5-linux-x64-v2/lib* /usr/local/cuda/lib64/
sudo cp cudnn-6.5-linux-x64-v2/cudnn.h /usr/local/cuda/include/
sudo reboot~/.bashrc
export CUDA_HOME=/usr/local/cuda
export CUDA_ROOT=/usr/local/cuda
export PATH=$PATH:$CUDA_ROOT/bin
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$CUDA_ROOT/lib64install Bazel 0.1.4, JAVA
sudo add-apt-repository -y ppa:webupd8team/java
sudo apt-get update
# Hack to silently agree license agreement
echo debconf shared/accepted-oracle-license-v1-1 select true | sudo debconf-set-selections
echo debconf shared/accepted-oracle-license-v1-1 seen true | sudo debconf-set-selections
sudo apt-get install -y oracle-java8-installer
#install bazel
sudo apt-get install pkg-config zip g++ zlib1g-dev
https://github.com/bazelbuild/bazel/releases/download/0.1.4/bazel-0.1.4-installer-linux-x86_64.sh
chmod +x bazel-0.1.4-installer-linux-x86_64.sh
./bazel-0.1.4-installer-linux-x86_64.sh --user
rm bazel-0.1.4-installer-linux-x86_64.shinstall tensorflow
git clone --recurse-submodules https://github.com/tensorflow/tensorflow
cd tensorflowbuild TensorFlow with GPU support using CUDA version 3.0 (currently required on AWS) via the unofficial settings.
TF_UNOFFICIAL_SETTING=1 ./configure
#type 3.0 to use CUDA version 3.0build
bazel build -c opt --config=cuda //tensorflow/cc:tutorials_example_trainer
bazel build -c opt --config=cuda //tensorflow/tools/pip_package:build_pip_package
bazel-bin/tensorflow/tools/pip_package/build_pip_package /tmp/tensorflow_pkg
sudo pip install --upgrade /tmp/tensorflow_pkg/tensorflow-0.6.0-cp27-none-linux_x86_64.whlverify working
python ~/tensorflow/tensorflow/models/image/mnist/convolutional.pyimport tensorflow as tf
tf_session = tf.Session()
x = tf.constant(1)
y = tf.constant(1)
tf_session.run(x + y)sudo pip install numpy scipy pandas sklearn matplotlib Pillow jupyter
sudo pip install keras
#version 1.1.1 oct 31st, 2016
sudo pip install keras_diagram
#from keras_diagram import ascii
#print(ascii(model))
sudo pip install kerasvis
#https://github.com/neuralyzer/kerasvis
#from kerasvis import DBLogger
#logger = DBLogger(comment="An example run")Install Instrtuctions:
https://github.com/gw0/docker-keras
Run with:
#python3 version
nvidia-docker run -it -p 8888:8888 -p 6006:6006 -v /home/jcx9dy/proj/cs6501:/mnt/cs6501/ -w /mnt/cs6501 gw000/keras:1.1.1-py3-tf-gpu bash
apt-get update
apt-get install vim less wget
#install jupyter
pip3 install --upgrade pip
pip3 install --upgrade numpy scipy pandas sklearn matplotlib Pillow jupyter
ipython
from notebook.auth import passwd
passwd()
'sha1:bcd259ccf...<your hashed password here>'
exit()
jupyter notebook --generate-config
vim ~/.jupyter/jupyter_notebook_config.py
# Set ip to '*' to bind on all interfaces (ips) for the public server
c.NotebookApp.ip = '*'
c.NotebookApp.password = u'sha1:bcd259ccf...<your hashed password here>'
c.NotebookApp.open_browser = False
c.NotebookApp.port = 8888
#start jupyter
jupyter notebook
#localhost:8888
#commit docker image
nvidia-docker commit -m "installed jupyter notebook and some python modules" -a "jcx9dy" <container id> jfx319/cs6501:keras1.1.1-py3-tf0.10-cuda7.5-cudnn5.1
#for convenience also add 'latest' tag
docker tag <image id> jfx319/cs6501:latest
#optionally, push to public repo (but keep in mind jupyter pw/config will be exposed)
docker login
docker push
## Test new image
nvidia-docker run -it -p 8888:8888 -p 6006:6006 -v /home/jcx9dy/proj/cs6501:/mnt/cs6501/ -w /mnt/cs6501 jfx319/cs6501:latest bash
##unstable if i just run "jupyter notebook" detached from docker cmd; more stable if i run bash first interactively starting jupyter notebook and then detaching from the docker container.
upgrade to latest keras (it changes every week!!):
sudo pip install git+git://github.com/fchollet/keras.git --upgradehack keras to only handle grayscale 16bit:
/usr/local/lib/python3.5/dist-packages/keras/preprocessing/image.py
if grayscale:
img = img.convert('L')
else: # Ensure 3 channel even when loaded image is grayscale
img = img.convert('RGB')
if target_size:
img = img.resize((target_size[1], target_size[0]))
return img if grayscale:
img = img.convert('I') #changed from 'L' to support 16bit grayscale
else:
img = img.convert('I') #still need 'I' b/c 'RGB' assumes 8bit
if target_size:
img = img.resize((target_size[1], target_size[0]))
if not grayscale:
x = np.asarray(img, dtype='float32')
img = np.dstack( (x,)*3 )
return imgreplaces mode='L' default 8bit with mode='I' for 16bit support (keras-team/keras#4486)
Also duplicate 16bit grayscale across 3 channels when using mode 'rgb'
jupyter notebook password is set to instance id
ec2metadata --instance-id
if don't want this already running one:
kill $(pgrep jupyter)
switch to screen, location of choice, and rerun jupyter notebookDownload weights from existing keras models (https://keras.io/applications/)
#Source: https://keras.io/applications/
#Weights are downloaded automatically when instantiating a model. They are stored at ~/.keras/models/
#Helper functions worth checking out in the github repo:
from keras.preprocessing import image
from keras.applications.resnet50 import preprocess_input, decode_predictions
from keras.applications.vgg16 import preprocess_input
from keras.applications.resnet50 import ResNet50
model = ResNet50(weights='imagenet')
from keras.applications.vgg16 import VGG16
model = VGG16(weights='imagenet')
from keras.applications.vgg19 import VGG19
model = VGG19(weights='imagenet')
from keras.applications.inception_v3 import InceptionV3
model = InceptionV3(weights='imagenet')
from keras.applications.xception import Xception
model = Xception(weights='imagenet')
from keras.applications.music_tagger_crnn import MusicTaggerCRNN
model = MusicTaggerCRNN(weights='msd')
# 528 MB vgg16_weights_tf_dim_ordering_tf_kernels.h5
# 549 MB vgg19_weights_tf_dim_ordering_tf_kernels.h5
# 91 MB inception_v3_weights_tf_dim_ordering_tf_kernels.h5
# 88 MB xception_weights_tf_dim_ordering_tf_kernels.h5
# 99 MB resnet50_weights_tf_dim_ordering_tf_kernels.h5
# 1.6 MB music_tagger_crnn_weights_tf_kernels_tf_dim_ordering.h5
https://fullstacknotes.com/mount-aws-s3-bucket-to-ubuntu-file-system/