initial release tested with Keras 2.1.2

README.md

@@ -1,33 +1,37 @@
-# Modified notebooks and Python files for Keras 2 and Python 3 from the fast.ai Deep Learning course
+# Modified notebooks and Python files for Keras 2 and Python 3 from the fast.ai Deep Learning course v.1
 The repository includes modified copies of the original Jupyter notebooks and Python files from the excellent
-(and really unique) deep learning course "Practical Deep Learning For Coders" Part 1 and Part 2,
+(and really unique) deep learning course "Practical Deep Learning For Coders" Part 1 and Part 2, v.1,
 created by [fast.ai](http://fast.ai). The [original files](https://github.com/fastai/courses)
 require Keras 1.
 
+The current version of the repository has been tested with **_Keras 2.1.2_**.
+The previous version, tested with _Keras 2.0.6_, is available [here](https://github.com/roebius/deeplearning_keras2/releases).
 ### Part 1
-Located in the _nbs_ folder. Tested with Keras 2.0.6 on both Ubuntu 16.04 and Python 3.5 (installed through apt-get) and
-MacOS 10.12.4 with Python 3.6 (installed with Homebrew). In Part 1 the Theano backend for Keras has been used. 
+Located in the _nbs_ folder. Tested on _Ubuntu 16.04_ and _Python 3.5_, installed through [Anaconda](https://www.anaconda.com), using the [Theano](http://deeplearning.net/software/theano/) 1.0.1 backend.  
 
 ### Part 2
-Located in the _nbs2_ folder. Tested with Keras 2.0.6 on Ubuntu 16.04 with Python 3.5 (installed through apt-get). In Part 2 the TensorFlow backend for Keras has been used.  
+Located in the _nbs2_ folder. Tested on _Ubuntu 16.04_ and _Python 3.5_, installed through [Anaconda](https://www.anaconda.com), using the [TensorFlow](https://www.tensorflow.org/) 1.3.0 backend.
+A few modules requiring PyTorch were also tested, using [PyTorch](http://pytorch.org/) 0.3.0.  
 
-The files keras.json.for\_TensorFlow and keras.json.for\_Theano provide a template for the appropriate keras.json file, based on which one of the two backends needs to be used.
+The files _keras.json.for\_TensorFlow_ and _keras.json.for\_Theano_ provide a template for the appropriate _keras.json_ file, based on which one of the two backends needs to be used by Keras.
 
-A Python 3 virtualenv has been used for both parts. In order to facilitate the installation of the required Python packages, this repository includes 
-also the requirement files that can be used with the pip command. These files include additional packages that might be useful for further exploration.  
+An _environment.yml_ file for creating a suitable [conda environment](https://conda.io/docs/user-guide/tasks/manage-environments.html) is provided. 
 
-The comments that I inserted in the modules generally start with *"# -"* when they are not just *"# Keras 2"*.
 
-### Notes about Part 2
-#### Issues 
+My goal has been to modify the original files to the minimum extent possible.The comments that I inserted in the modules generally start with *"# -"* when they are not just *"# Keras 2"*.
+
+### Notes and issues about Part 2
+*neural-style.ipynb*: due to a function parameter change in _Keras 2.1_, the _VGG16_ provided by _Keras 2.1_ has been used instead of the original custom module _vgg16\_avg.py_
+
 *rossman.ipynb*: section "Using 3rd place data" has been left out for lack of the required data
 
-*spelling_bee_RNN.ipynb*: after the main part of the notebook, in the final "Test code ..." section I was not able to solve an issue with the K.conv1d cell not working 
+*spelling_bee_RNN.ipynb* and *attention_wrapper.py*: due to the changed implementation of the recurrent.py module in Keras 2.1, the attention part of the notebook doesn't work anymore
 
 *taxi_data_prep_and_mlp.ipynb*: section "Uh oh ..." has been left out. Caveat: running all the notebook at once exhausted 128 GB RAM; I was able to run each section individually only after resetting the notebook kernel each time
 
 *tiramisu-keras.ipynb*: in order to run the larger size model I had to reset the notebook kernel in order to free up enough GPU memory (almost 12 GB) and jump directly to the model
 
+
 #### Left-out modules
 *neural-style-pytorch.ipynb* (found no way to load the VGG weights; it looks like some version compatibility issue)
 

environment.yml

@@ -0,0 +1,185 @@
+name: p3
+channels:
+  - pytorch
+  - conda-forge
+  - defaults
+dependencies:
+  - backports.weakref=1.0rc1=py35_0
+  - bleach=1.5.0=py35_0
+  - distributed=1.20.2=py35_0
+  - html5lib=0.9999999=py35_0
+  - jupyter_contrib_core=0.3.3=py35_1
+  - jupyter_nbextensions_configurator=0.3.0=py35_0
+  - markdown=2.6.9=py35_0
+  - asn1crypto=0.23.0=py35h4ab26a5_0
+  - backports=1.0=py35hd471ac7_1
+  - bcolz=1.1.2=py35hcb27967_0
+  - binutils_impl_linux-64=2.28.1=h04c84fa_2
+  - binutils_linux-64=7.2.0=25
+  - bokeh=0.12.13=py35h2f9c1c0_0
+  - boto=2.48.0=py35h2cfd601_1
+  - bz2file=0.98=py35_0
+  - bzip2=1.0.6=h6d464ef_2
+  - ca-certificates=2017.08.26=h1d4fec5_0
+  - certifi=2017.11.5=py35h9749603_0
+  - cffi=1.11.2=py35hc7b2db7_0
+  - chardet=3.0.4=py35hb6e9ddf_1
+  - click=6.7=py35h353a69f_0
+  - cloudpickle=0.5.2=py35hbe86bc5_0
+  - cryptography=2.1.4=py35hbeb2da1_0
+  - cudatoolkit=8.0=3
+  - cudnn=6.0.21=cuda8.0_0
+  - cycler=0.10.0=py35hc4d5149_0
+  - cython=0.27.3=py35h6cdc64b_0
+  - dask=0.16.0=py35hcb8ecc8_0
+  - dask-core=0.16.0=py35hfc66869_0
+  - dbus=1.10.22=h3b5a359_0
+  - decorator=4.1.2=py35h3a268aa_0
+  - entrypoints=0.2.3=py35h48174a2_2
+  - expat=2.2.5=he0dffb1_0
+  - fastcache=1.0.2=py35hec2bbaa_0
+  - fontconfig=2.12.4=h88586e7_1
+  - freetype=2.8=hab7d2ae_1
+  - gcc_impl_linux-64=7.2.0=hc5ce805_2
+  - gcc_linux-64=7.2.0=25
+  - gensim=3.1.0=py35h7300b16_0
+  - glib=2.53.6=h5d9569c_2
+  - gmp=6.1.2=h6c8ec71_1
+  - gmpy2=2.0.8=py35hd0a1c9a_2
+  - gst-plugins-base=1.12.2=he3457e5_0
+  - gstreamer=1.12.2=h4f93127_0
+  - gxx_impl_linux-64=7.2.0=hd3faf3d_2
+  - gxx_linux-64=7.2.0=25
+  - h5py=2.7.1=py35h8d53cdc_0
+  - hdf5=1.10.1=h9caa474_1
+  - heapdict=1.0.0=py35h51e6c10_0
+  - icu=58.2=h9c2bf20_1
+  - idna=2.6=py35h8605a33_1
+  - imageio=2.2.0=py35hd0a6de2_0
+  - intel-openmp=2018.0.0=hc7b2577_8
+  - ipykernel=4.7.0=py35h2f9c1c0_0
+  - ipython=6.2.1=py35hd850d2a_1
+  - ipython_genutils=0.2.0=py35hc9e07d0_0
+  - ipywidgets=7.0.5=py35h8147dc1_0
+  - jedi=0.11.0=py35_2
+  - jinja2=2.10=py35h480ab6d_0
+  - jpeg=9b=h024ee3a_2
+  - jsonschema=2.6.0=py35h4395190_0
+  - jupyter=1.0.0=py35hd38625c_0
+  - jupyter_client=5.1.0=py35h2bff583_0
+  - jupyter_console=5.2.0=py35h4044a63_1
+  - jupyter_core=4.4.0=py35ha89e94b_0
+  - keras=2.1.2=py35_0
+  - libedit=3.1=heed3624_0
+  - libffi=3.2.1=hd88cf55_4
+  - libgcc=7.2.0=h69d50b8_2
+  - libgcc-ng=7.2.0=h7cc24e2_2
+  - libgfortran-ng=7.2.0=h9f7466a_2
+  - libgpuarray=0.7.5=h14c3975_0
+  - libpng=1.6.32=hbd3595f_4
+  - libprotobuf=3.4.1=h5b8497f_0
+  - libsodium=1.0.15=hf101ebd_0
+  - libstdcxx-ng=7.2.0=h7a57d05_2
+  - libtiff=4.0.9=h28f6b97_0
+  - libxcb=1.12=hcd93eb1_4
+  - libxml2=2.9.4=h2e8b1d7_6
+  - locket=0.2.0=py35h170bc82_1
+  - lzo=2.10=h49e0be7_2
+  - mako=1.0.7=py35h69899ea_0
+  - markupsafe=1.0=py35h4f4fcf6_1
+  - matplotlib=2.1.1=py35ha26af80_0
+  - mistune=0.8.1=py35h9251d8c_0
+  - mkl=2018.0.1=h19d6760_4
+  - mkl-service=1.1.2=py35h0fc7090_4
+  - mpc=1.0.3=hec55b23_5
+  - mpfr=3.1.5=h11a74b3_2
+  - mpmath=1.0.0=py35h7ce6e34_2
+  - msgpack-python=0.4.8=py35h783f4c8_0
+  - nbconvert=5.3.1=py35hc5194e3_0
+  - nbformat=4.4.0=py35h12e6e07_0
+  - ncurses=6.0=h9df7e31_2
+  - networkx=2.0=py35hc690e10_0
+  - nltk=3.2.5=py35h09ad193_0
+  - notebook=5.2.2=py35he644770_0
+  - numexpr=2.6.4=py35h119f745_0
+  - numpy=1.13.3=py35hd829ed6_0
+  - olefile=0.44=py35h2c86149_0
+  - openssl=1.0.2n=hb7f436b_0
+  - pandas=0.22.0=py35hf484d3e_0
+  - pandoc=1.19.2.1=hea2e7c5_1
+  - pandocfilters=1.4.2=py35h1565a15_1
+  - parso=0.1.1=py35h1b200a3_0
+  - partd=0.3.8=py35h68187f2_0
+  - pcre=8.41=hc27e229_1
+  - pexpect=4.3.0=py35hf410859_0
+  - pickleshare=0.7.4=py35hd57304d_0
+  - pillow=5.0.0=py35h3deb7b8_0
+  - pip=9.0.1=py35h7e7da9d_4
+  - prompt_toolkit=1.0.15=py35hc09de7a_0
+  - protobuf=3.4.1=py35he6b9134_0
+  - psutil=5.4.1=py35h2e39a06_0
+  - ptyprocess=0.5.2=py35h38ce0a3_0
+  - pycparser=2.18=py35h61b3040_1
+  - pygments=2.2.0=py35h0f41973_0
+  - pygpu=0.7.5=py35h14c3975_0
+  - pyopenssl=17.5.0=py35h4f8b8c8_0
+  - pyparsing=2.2.0=py35h041ed72_1
+  - pyqt=5.6.0=py35h0e41ada_5
+  - pysocks=1.6.7=py35h6aefbb0_1
+  - pytables=3.4.2=py35hfa98db7_2
+  - python=3.5.4=h417fded_24
+  - python-dateutil=2.6.1=py35h90d5b31_1
+  - pytz=2017.3=py35hb13c558_0
+  - pywavelets=0.5.2=py35h53ec731_0
+  - pyyaml=3.12=py35h46ef4ae_1
+  - pyzmq=16.0.3=py35ha889422_0
+  - qt=5.6.2=h974d657_12
+  - qtconsole=4.3.1=py35h4626a06_0
+  - readline=7.0=ha6073c6_4
+  - requests=2.18.4=py35hb9e6ad1_1
+  - scikit-image=0.13.1=py35h14c3975_1
+  - scikit-learn=0.19.1=py35hbf1f462_0
+  - scipy=1.0.0=py35hcbbe4a2_0
+  - setuptools=36.5.0=py35ha8c1747_0
+  - simplegeneric=0.8.1=py35h2ec4104_0
+  - sip=4.18.1=py35h9eaea60_2
+  - six=1.11.0=py35h423b573_1
+  - smart_open=1.5.3=py35_0
+  - sortedcontainers=1.5.7=py35h683703c_0
+  - sqlite=3.20.1=hb898158_2
+  - sympy=1.1.1=py35h919b29a_0
+  - tblib=1.3.2=py35hf1eb0b4_0
+  - tensorflow=1.3.0=0
+  - tensorflow-base=1.3.0=py35h79a3156_1
+  - tensorflow-gpu=1.3.0=0
+  - tensorflow-gpu-base=1.3.0=py35cuda8.0cudnn6.0_1
+  - tensorflow-tensorboard=0.1.5=py35_0
+  - terminado=0.6=py35hce234ed_0
+  - testpath=0.3.1=py35had42eaf_0
+  - theano=1.0.1=py35h6bb024c_0
+  - tk=8.6.7=hc745277_3
+  - toolz=0.8.2=py35h90f1797_0
+  - tornado=4.5.2=py35hf879e1d_0
+  - tqdm=4.19.4=py35h68e51d2_0
+  - traitlets=4.3.2=py35ha522a97_0
+  - ujson=1.35=py35_0
+  - urllib3=1.22=py35h2ab6e29_0
+  - wcwidth=0.1.7=py35hcd08066_0
+  - webencodings=0.5.1=py35hb6cf162_1
+  - werkzeug=0.12.2=py35hbfc1ea6_0
+  - wheel=0.30.0=py35hd3883cf_1
+  - widgetsnbextension=3.0.8=py35h84cb72a_0
+  - xz=5.2.3=h55aa19d_2
+  - yaml=0.1.7=had09818_2
+  - zeromq=4.2.2=hbedb6e5_2
+  - zict=0.1.3=py35h29275ca_0
+  - zlib=1.2.11=ha838bed_2
+  - pytorch=0.3.0=py35_cuda8.0.61_cudnn7.0.3hb362f6e_4
+  - torchvision=0.2.0=py35heaa392f_1
+  - pip:
+    - keras-tqdm==2.0.1
+    - tables==3.4.2
+    - torch==0.3.0.post4
+    - xgboost==0.7.post3
+prefix: /home/roebius/anaconda/envs/f1
+