Repository navigation
release 2017-02 follow-up #516
Description
Activity
audit-trail:
- remove nose from default build (unless a package requirement still pulls it),
- infrastructure:
- separate 32bit from 64bit audit trails (32bit is dying, changelogs are becoming different, had to do it by hand for 2017-01):
- WinPythonQt5-64bit-3.6.1.1.md summary
- WinPythonQt5-64bit-3.6.1.1_History.md changelog
- pyqt-5.8.2
- scikit-image-0.13.0
- Jupyterlab-0.19.0, qtconsole-3.0.0, ipykernel-4.6.1
- numba-0.32.0
- lmfit-0.9.6
- WinPythonZero : joblib, dask[bag, delayed], brotli, python-snappy
- fastparquet-0.0.5
- holoview-1.7.0rc1 (still labelled 1.7.dev8... a string bug)
- ipython-6.0.0rc1
- spyder-3.1.4dev0-20170415
- infrastructure : switch from winpython.qt to winpython._vendor.Qtpy (drop specifics to allow pyside2)
- pylint-1.7.1
- thriftpy-0.3.9+ cythonized thanks to cgohlke
- backpedal to tornado-4.4.3 (windows: No supported color terminal library tornadoweb/tornado#2013)
(http://stackoverflow.com/questions/43443974/no-supported-color-terminal-library-python-jupyter) - tornado-4.5.1
- spyder-3.1.4dev0-20170421 (just missing translations)
- tensorflow-1.1.0
- jupyterlab-0.20.0, jupyter-launcher
- pandas-0.20.0rc1
- remove MPL from bokeh example (ongoing deprecation of bokeh-mpl link)
- an icon jupyterlab shall be usefull in this intermediate times
build1 (2017-04-23: ipython-6.0, holoviews-1.7, matplotlib-2.0.1, droping python 2.7/3.4 as we look for C++11 packages)
- holoviews-1.7.0
- ipython-6.0.0
- jupyterlab_launcher
- Jupyterlab-0.20.3
- Edward-1.3.0
- spyder-3.1.4
- 2017-04-29 fail test on Vitables for Qt5 (even on WinPython-64bit-3.5.3.0Qt5)
- vitables-2.2.1dev0 (with QtPy patch)
- feather_format-0.3.1
- matplotlib-2.0.1
- pandas-0.20.0rc2
- pandas-0.20.1
- llvmlite-0.18.0, numba-0.33
- dask-0.14.3 (pandas-0.20.1 compta.)
- supersmoother
- stay on nbdime-0.2 as nbdime-0.3 requires gitpython which requires gitdb2 which requires smmap , all requiring pygit ... a bit too much at the moment
- mypy-0.510
build2 (2017-05-07: vitables, matplotlib-2.0.1)
- jupyterlab-0.21.0,
- matplotlib-2.0.2
- pyarrow-0.3.0 on python-3.6 64bits
- jupyterlab-0.22.0
- pyarrow-0.4.0
- seaborn-0.8.dev0-2070512 (pandas-0.20 fix)
- jupyterlab-0.22.1
- FAILED try to upgrade jedi-0.10.3dev21070525 (works less nicely than jedi-0.10.2 on spyder editor)
- seaborn-0.8.dev0-2070522
- nbdime-0.3.0 (as even Windows is on GIT )
- spyder-3.2.0dev-20170524
- imageio-2.2.0
- pybind11
- ipython-6.1.0 (announced for may 24th)
- jupyterlab-0.23.1
- pandas-0.20.2
- jupyterlab-0.24.1
- spyder-3.1.2dev20160615
- pyarrow-0.4.0 (python-3.6 64 bit, with parquet support )
- bokeh-0.12.6
- pyarrow-0.4.1 (python-3.5/3.6 64 bit, with parquet support )
- holoviews-0.18.2dev-20160617
- pygame (there is also some non-game use of it, even with jupyter notebook https://www.youtube.com/watch?v=phtDJYpceqw)
build3 (2017-06-18, delayed by a busy week-end: ipython-6.1, pandas-0.20.2, jupyterlab-0.23.1, spyder-3.2dev )
- scikit-learn-0.18.2 (compatibility patch with numpy-0.13)
- check cvxpy possibility (initial fail, pyreadline is a pain)
- holoviews 1.8.0
- python-3.6.2rc2
- holoviews 1.8.1
- jupyterlab-0.25.1
- pyreadline i hate you
- pandas-0.20.3
- seaborn-0.8.0
- cvxpy-0.4.10
- spyder-3.2-20170709am (3.2 final maybe)
- keras_vis-0.4.1
- upgrade to vc_redist_v3 (https://www.microsoft.com/en-us/download/details.aspx?id=53840
, build 14.0.24215), required for some future package - plotnine replaces ggpy
- feather_format-0.4 ... will have dependancy on pyarrow (aye aye: no wheel yet)
b4 (2017-07-08, may slip by 7 days more)
- numpy-0.13.1
- mpldatacursor
- spyder-notebook
b5 (2017-07-16, for the failed b4 attempt on python-3.6.2rc2 32bit)
- python-3.6.2 final
- typed-ast-1.0.4 (and not higher)
- mpl-scatter-density
- Jupyterlab widgets extension requires npm / nodejs-6.11.1
- Jupyterlab-0.26.4
- altair-widgets
- spyder-3.2.1.dev20170729
- spyder-reports
- spyder-terminal
b6 (2017-07-30, nodejs, spyder-terminal)
- python-3.6.2 final (was still python-3.6.2rc2)
- ipywidgets-7.0.0b3 (should be compatible with jupyterlab-0.26.4, so we should be see what jupyterlab truly is)
- Pulp-1.6.8
- datashader
- holoviews-1.8.2
- spyder-3.2.1.dev20170805
- notebook-5.1.0rc1
- ffmpeg, moviepy
- ipywidgets-7.0.0b7 (apparently ipyleaftlet/bqplot/bokeh will work on classic notebook, not in jupyterlab before a while)
- Jupyterlab-0.26.5
- python-3.5.4
- notebook-5.1rc1
- back to ipywidgets-6.0.0: ipywidgets-7 with a big and stable ecosystem is looking still month away
rc (2017-08-13) : (ffmpeg)
may drop it (unless we slip again of 7 days) :
- fresh scikit-learn-0.19.0 has some compatibility breakage with scikit-optimize
- altair-1.3
next time:
- baking ipywidget-7.0betas still baking, and so ecosystem not compatible yet (wait in 1 month)
- interesting "categorical" bokeh-0.12.7 may arrive after August 22nd
- altair-2.0
experiments:
- not turbodbc-2.1 (... too invasive for now).
- rely on qtpy for Qt interface (should allow Pyside2, one day)
- cntk-2.0 is now officially python-3.6 (33 Mo yet, because of opencv which adds 13 Mo, then add MSMpiSetup.exe for 5 Mo) . the need to install MSMpisetup.exe is a problem for WinPython case. KEras support supposed to work .... wiating to see benchmark on cpu
- review nddime-0.3 consequences
- (long term idea) remove heavy PyQt5 need from wppm:
- idle gui: painfull
- Flask: very painfull
- there may be other things.
- added this to the milestone on Mar 26, 2017
Would it be possible to add GPflow too?
I haven't succeeded in installing it yet :(llvmlite updated to v0.17.1 http://www.lfd.uci.edu/~gohlke/pythonlibs/#llvmlite
build1 (2017-04-2x): numpy-0.11.2, pandas-0.19.0, bokeh-0.12.5
main features (since WinPython 2017-01 ):
- pandas-0.20.0rc1, fastparquet, spyder-3.1.4dev0-20170421 (one commit from release)
- Jupyterlab-0.20.0 (still alpha), add "Jupyter Lab" icon, remove "QtLinguist" icon)
- ipython-6.0.0
Areas of interest for testers:
- pandas-0.20, spyder-3.1.4, ipython-6.0 (stability ?)
- Jupyterlab (alpha)
Next Focus:
- tweaks around Jupyterlab / Web technologies
- tweaks around Pandas / Parquet / Arrow / Data technologies
- hope for numpy-1.12.2 ?
MD5 SHA-1 SHA-256 Binary Size ddd8af87f02530e61f39d61a247b70b8 383822c18b5a8cef964071d419f2fb3cc278e2c5 a4a5bfcec76c05c9326388f22a72012d7c00727b97e67972dc9bdc2e716d54f8 WinPython-64bit-3.6.1.1Zerob1.exe 25 389 938 Bytes 40714b682e1485daf4b098d2f5179ccf 0abe7e01eff0a45f5ee87173db8a0b27251ca6bf b5e745379c024b581ed0b3a10ce1ae00203f04534e3373156e93be3a2c7091df WinPython-32bit-3.6.1.1Zerob1.exe 24 514 948 Bytes 7e5a039129d393b60079b5a4901176a5 ea9b3ced0907dc5f9648c93e5a95af808b4cdc88 abe34721162a86f0e8948eefdf48328087f9286b79cc5ce346bfe422189b6e21 WinPython-32bit-3.6.1.1Qt5b1.exe 276 390 069 Bytes eddf491ce62b2122b58c092329345b16 03dad07b86c5f3e0998c2d5a95d9b54579fcd510 0a3137f52e0a7f8a1b48ed679e823da0f6098aab35bd4a5e01990056f7cf9ec8 WinPython-64bit-3.6.1.1Qt5b1.exe 331 560 104 Bytes History of changes for WinPython-64bit 3.6.1.1Zero
The following changes were made to WinPython-64bit distribution since version 3.6.1.0Zero.
Python packages
New packages:
-
brotli 0.6.0 (Python binding of the Brotli compression library)
-
cloudpickle 0.2.2 (Extended pickling support for Python objects)
-
dask 0.14.1 (Minimal task scheduling abstraction)
-
joblib 0.11 (Lightweight pipelining: using Python functions as pipeline jobs.)
-
locket 0.2.0 (File-based locks for Python for Linux and Windows)
-
partd 0.3.7 (Appendable key-value storage)
-
python_snappy 0.5 (Python library for the snappy compression library from Google)
-
toolz 0.8.2 (List processing tools and functional utilities)
Upgraded packages:
-
jedi 0.10.0 → 0.10.2 (An autocompletion tool for Python that can be used for text editors)
-
prompt_toolkit 1.0.13 → 1.0.14 (Library for building powerful interactive command lines in Python)
-
ptpython 0.36 → 0.39 (Python REPL build on top of prompt_toolkit)
-
pyodbc 4.0.15 → 4.0.16 (DB API Module for ODBC)
-
setuptools 34.3.2 → 35.0.1 (Download, build, install, upgrade, and uninstall Python packages - easily)
-
winpython 1.7.20170128 → 1.9.20170423 (WinPython distribution tools, including WPPM (package manager))
History of changes for WinPython-64bit 3.6.1.1Qt5
The following changes were made to WinPython-64bit distribution since version 3.6.1.0Qt5.
Python packages
New packages:
-
brotli 0.6.0 (Python binding of the Brotli compression library)
-
fastparquet 0.0.5 (Python support for Parquet file format)
-
jupyterlab_launcher 0.1.4 (Jupyter Launcher)
-
ply 3.10 (Python Lex & Yacc)
-
python_snappy 0.5 (Python library for the snappy compression library from Google)
-
thriftpy 0.3.9 (Pure python implementation of Apache Thrift.)
Upgraded packages:
-
astroid 1.4.9 → 1.5.2 (Rebuild a new abstract syntax tree from Python's ast (required for pylint))
-
blosc 1.5.0 → 1.5.1 (Blosc data compressor)
-
bokeh 0.12.4 → 0.12.5 (Statistical and novel interactive HTML plots for Python)
-
certifi 2017.1.23 → 2017.4.17 (Python package for providing Mozilla's CA Bundle.)
-
chardet 2.3.0 → 3.0.2 (Universal encoding detector for Python 2 and 3)
-
db.py 0.5.2 → 0.5.3 (a db package that doesn't suck)
-
flask 0.12 → 0.12.1 (A microframework based on Werkzeug, Jinja2 and good intentions)
-
flask_cors 2.1.2 → 3.0.2 (A Flask extension adding a decorator for CORS support)
-
ipykernel 4.5.2 → 4.6.1 (IPython Kernel for Jupyter)
-
ipython 5.3.0 → 6.0.0 (Enhanced Python shell)
-
jedi 0.10.0 → 0.10.2 (An autocompletion tool for Python that can be used for text editors)
-
jinja2 2.9.5 → 2.9.6 (Sandboxed template engine (provides a Django-like non-XML syntax and compiles templates into executable python code))
-
jupyter_client 5.0.0 → 5.0.1 (Jupyter protocol implementation and client libraries)
-
jupyterlab 0.18.1 → 0.20.0 (Jupyter lab environment notebook server extension)
-
keras 2.0.2 → 2.0.3 (Theano-based Deep Learning library)
-
llvmlite 0.16.0 → 0.17.1 (lightweight wrapper around basic LLVM functionality)
-
lmfit 0.9.5 → 0.9.6 (Least-Squares Minimization with Bounds and Constraints)
-
notebook 5.0.0rc2 → 5.0.0 (# Jupyter Notebook)
-
numba 0.31.0 → 0.32.0 (compiling Python code using LLVM)
-
oct2py 4.0.5 → 4.0.6 (Python to GNU Octave bridge --> run m-files from python.)
-
pandas 0.19.2 → 0.20.0rc1 (Powerful data structures for data analysis, time series and statistics)
-
pillow 4.0.0 → 4.1.1 (Python Imaging Library (fork))
-
prompt_toolkit 1.0.13 → 1.0.14 (Library for building powerful interactive command lines in Python)
-
psutil 5.2.0 → 5.2.2 (Provides an interface for retrieving information on all running processes and system utilization (CPU, disk, memory, network) in a portable way)
-
ptpython 0.36 → 0.39 (Python REPL build on top of prompt_toolkit)
-
pylint 1.6.5 → 1.7.1 (Logilab code analysis module: analyzes Python source code looking for bugs and signs of poor quality)
-
pyodbc 4.0.15 → 4.0.16 (DB API Module for ODBC)
-
pyqt5 5.8.1 → 5.8.2 (Python bindings for the Qt cross platform GUI toolkit)
-
pytz 2016.10 → 2017.2 (World Timezone Definitions for Python)
-
pywin32 220.1 → 221 (Python library for Windows)
-
qtconsole 4.2.1 → 4.3.0 (Jupyter Qt console)
-
rx 1.5.8 → 1.5.9 (Reactive Extensions (Rx) for Python)
-
scikit_image 0.13.dev0 → 0.13.0 (Image processing toolbox for SciPy)
-
setuptools 34.3.2 → 35.0.1 (Download, build, install, upgrade, and uninstall Python packages - easily)
-
sip 4.19.1 → 4.19.2 (Python extension module generator for C and C++ libraries)
-
sortedcollections 0.4.2 → 0.5.3 (Python Sorted Collections)
-
sphinx 1.5.2 → 1.5.5 (Tool for generating documentation which uses reStructuredText as its markup language)
-
sqlalchemy 1.1.6 → 1.1.9 (SQL Toolkit and Object Relational Mapper)
-
tables 3.3.0 → 3.4.2 (Package based on HDF5 library for managing hierarchical datasets (extremely large amounts of data))
-
tblib 1.3.0 → 1.3.2 (Traceback serialization library.)
-
tensorflow 1.0.1 → 1.1.0 (TensorFlow helps the tensors flow)
-
tornado 4.4.2 → 4.5.1 (Scalable, non-blocking web server and tools (required for IPython notebook))
-
webencodings 0.5 → 0.5.1 (Character encoding aliases for legacy web content)
-
winpython 1.7.20170128 → 1.9.20170423 (WinPython distribution tools, including WPPM (package manager))
-
xarray 0.9.1 → 0.9.5 (N-D labeled arrays and datasets in Python)
Removed packages:
- nose 1.3.7 (nose is a discovery-based unittest extension (e.g. NumPy test module is using nose))
matplotlib v2.0.1 is available today. See http://www.lfd.uci.edu/~gohlke/pythonlibs/#matplotlib
Please include Python 3.5 for future test builds.
llvmlite‑0.18.0 is available. May be required for numba-0.33.0.
build2 (2017-05-07): pandas-0.20.1, matplotlib-2.0.1, spyder-3.1.4
main features (since WinPython 2017-02 build1):
- pandas-0.20.1, matplotlib-2.0.1, numba-0.33.0, dask-0.14.2, feather-format-0.3.1
- Vitables-2.2.1dev0 (PyQt5 compatible at last)
- ipython-6.0.0, spyder-3.1.4, mypy-0.510 (on Python-3.6)
Areas of interest for testers: (to contribute bug examples and patches)
- check compatibility of other packages with pandas-0.20.1
- help Vitables maintainer to stabilize next release
- help check numba-0.33 new features
Next Focus:
- hunt Jupyterlab / ipywidgets progress towards usable beta
- trying to get pyarrow working on Windows
- checking cgohlke's move (or not) on numpy > 1.11.3
MD5 SHA-1 SHA-256 Binary Size f47281fc6f0079ddbd02e664f94a3a73 3aeb698a6b7a2aa6f0c7b2c49ab355f9ab10febd ed4119b8d19099e4f74b832afeaea068b7cd1b265e3e0c6514fed47ec3e5e95e WinPython-32bit-3.5.3.2Qt5b2.exe 276 602 378 Bytes 717064ebc4b26847f98dc9d22304aad6 fe2b889e79e403b0c21940933e5b8c121a6629b5 638241733ddb01d8343d6d5418000b19630d6eadb8d80e4d4c5b562a265fead5 WinPython-64bit-3.5.3.2Qt5b2.exe 335 822 118 Bytes 38855a7192121af5d23c995ddb0a1f25 d5091bb959cba4479adbff5d29801c7f04a23ad4 3faa7e9d497f1e3a4489997e570cba1d636828997718df6313aa24d25e456e24 WinPython-64bit-3.5.3.2Zerob2.exe 25 049 967 Bytes 7b90b3d1f80ed62e6d9b35b8ba37abfb 4c8cc4dfe9921e1d3a1d7de1bcdf211791ac2dc9 b780dc187d53e7f272b7779411d9663b3ba9dd453324fb2a93f0e8cad1ea4e4b WinPython-32bit-3.5.3.2Zerob2.exe 24 274 607 Bytes 87c8394b9e79210bf4c60c60f7cf8281 3df5975a0f612b5c895a4226ae0db1aacd013701 7629ec8ae951e3977ab1475d5ad3c8027ae1955ee061b6872d2f9ec621bdfd81 WinPython-32bit-3.6.1.1Qt5b2.exe 276 867 380 Bytes 981639d25ecc5a5a2c2f0490c8be6361 39a331d1d1f69186826060d16978eeb8cf3e28c9 4c6199c884b349ec1e42c15f56dbfc4c1817ed475cd68a0603b2884a77b5a438 WinPython-64bit-3.6.1.1Qt5b2.exe 332 122 771 Bytes a9e989a602fdc843b1702852eb835f03 03fcac83ff953b32bbc1e3303fb964e8693088e8 58cc0e0abe5f3b9abd7ebcfc571f363e3951732c494029f04c218b31eb459300 WinPython-64bit-3.6.1.1Zerob2.exe 25 420 165 Bytes bbada1f86b975f4994df490d8994631c 2c54d447dd93470a7c902e445336d91fe763aa76 3f9864af39fb714691d5ce09558314fbb29e95172ade0e0d3dfa49e674eb1dc7 WinPython-32bit-3.6.1.1Zerob2.exe 24 551 778 Bytes e5a3d7f8b8af3608dd11ed24faad15ad 2551dedbdee3078c5132efc8b622438f0b2e8b88 f99d34f5486d513da97a8c70510a87621f6c3e019c12786fe733e8e86969f9eb WinPython-64bit-3.6.2.0Qt5b4.exe 341 753 866 Bytes History of changes for WinPython-64bit 3.5.3.2Qt5
The following changes were made to WinPython-64bit distribution since version 3.5.3.1Qt5.
Python packages
New packages:
-
brotli 0.6.0 (Python binding of the Brotli compression library)
-
fastparquet 0.0.6 (Python support for Parquet file format)
-
feather_format 0.3.1 (Python interface to the Apache Arrow-based Feather File Format)
-
jupyterlab 0.20.3 (Jupyter lab environment notebook server extension)
-
jupyterlab_launcher 0.1.4 (Jupyter Launcher)
-
ply 3.10 (Python Lex & Yacc)
-
python_snappy 0.5.1 (Python library for the snappy compression library from Google)
-
supersmoother 0.4 (Python implementation of Friedman's Supersmoother)
-
thriftpy 0.3.9 (Pure python implementation of Apache Thrift.)
-
vitables 2.2.1.dev0 (Graphical tool for browsing and editing files in both HDF5 and PyTables formats)
Upgraded packages:
-
astroid 1.4.9 → 1.5.2 (Rebuild a new abstract syntax tree from Python's ast (required for pylint))
-
beautifulsoup4 4.5.3 → 4.6.0 (Screen-scraping library)
-
blosc 1.5.0 → 1.5.1 (Blosc data compressor)
-
bokeh 0.12.4 → 0.12.5 (Statistical and novel interactive HTML plots for Python)
-
certifi 2017.1.23 → 2017.4.17 (Python package for providing Mozilla's CA Bundle.)
-
chardet 2.3.0 → 3.0.2 (Universal encoding detector for Python 2 and 3)
-
colorama 0.3.7 → 0.3.9 (Cross-platform colored terminal text)
-
dask 0.14.1 → 0.14.3 (Minimal task scheduling abstraction)
-
db.py 0.5.2 → 0.5.3 (a db package that doesn't suck)
-
distributed 1.16.1 → 1.16.2 (Distributed computing)
-
edward 1.2.4 → 1.3.1 (A library for probabilistic modeling, inference, and criticism. Deep generative models, variational inference. Runs on TensorFlow.)
-
flask 0.12 → 0.12.1 (A microframework based on Werkzeug, Jinja2 and good intentions)
-
flask_cors 2.1.2 → 3.0.2 (A Flask extension adding a decorator for CORS support)
-
holoviews 1.7.dev8 → 1.7.0 (Composable, declarative data structures for building complex visualizations easily.)
-
ipykernel 4.5.2 → 4.6.1 (IPython Kernel for Jupyter)
-
ipython 5.3.0 → 6.0.0 (Enhanced Python shell)
-
jedi 0.10.0 → 0.10.2 (An autocompletion tool for Python that can be used for text editors)
-
jinja2 2.9.5 → 2.9.6 (Sandboxed template engine (provides a Django-like non-XML syntax and compiles templates into executable python code))
-
jupyter_client 5.0.0 → 5.0.1 (Jupyter protocol implementation and client libraries)
-
keras 2.0.2 → 2.0.4 (Theano-based Deep Learning library)
-
lazy_object_proxy 1.2.2 → 1.3.1 (A fast and thorough lazy object proxy.)
-
llvmlite 0.16.0 → 0.18.0 (lightweight wrapper around basic LLVM functionality)
-
lmfit 0.9.5 → 0.9.6 (Least-Squares Minimization with Bounds and Constraints)
-
matplotlib 2.0.0 → 2.0.1 (2D plotting library (embeddable in GUIs created with PyQt))
-
notebook 5.0.0rc2 → 5.0.0 (# Jupyter Notebook)
-
numba 0.31.0 → 0.33.0 (compiling Python code using LLVM)
-
oct2py 4.0.5 → 4.0.6 (Python to GNU Octave bridge --> run m-files from python.)
-
pandas 0.19.2 → 0.20.1 (Powerful data structures for data analysis, time series and statistics)
-
param 1.5.0 → 1.5.1 (Declarative Python programming using Parameters.)
-
partd 0.3.7 → 0.3.8 (Appendable key-value storage)
-
pillow 4.0.0 → 4.1.1 (Python Imaging Library (fork))
-
prompt_toolkit 1.0.13 → 1.0.14 (Library for building powerful interactive command lines in Python)
-
protobuf 3.2.0 → 3.3.0 (Protocol Buffers - Google's data interchange format)
-
psutil 5.2.0 → 5.2.2 (Provides an interface for retrieving information on all running processes and system utilization (CPU, disk, memory, network) in a portable way)
-
ptpython 0.36 → 0.39 (Python REPL build on top of prompt_toolkit)
-
pylint 1.6.5 → 1.7.1 (Logilab code analysis module: analyzes Python source code looking for bugs and signs of poor quality)
-
pyodbc 4.0.15 → 4.0.16 (DB API Module for ODBC)
-
pyqt5 5.8.1 → 5.8.2 (Python bindings for the Qt cross platform GUI toolkit)
-
pytz 2016.10 → 2017.2 (World Timezone Definitions for Python)
-
pywin32 220.1 → 221 (Python library for Windows)
-
qtconsole 4.2.1 → 4.3.0 (Jupyter Qt console)
-
requests_file 1.4.1 → 1.4.2 (File transport adapter for Requests)
-
rx 1.5.8 → 1.5.9 (Reactive Extensions (Rx) for Python)
-
s3fs 0.0.9 → 0.1.0 (Convenient Filesystem interface over S3)
-
scikit_image 0.13.dev0 → 0.13.0 (Image processing toolbox for SciPy)
-
setuptools 34.3.2 → 35.0.2 (Download, build, install, upgrade, and uninstall Python packages - easily)
-
sip 4.19.1 → 4.19.2 (Python extension module generator for C and C++ libraries)
-
sortedcollections 0.4.2 → 0.5.3 (Python Sorted Collections)
-
sphinx 1.5.2 → 1.5.5 (Tool for generating documentation which uses reStructuredText as its markup language)
-
spyder 3.1.4.dev0 → 3.1.4 (Scientific PYthon Development EnviRonment: designed for interactive computing and data visualisation with a simple and intuitive user interface)
-
sqlalchemy 1.1.6 → 1.1.9 (SQL Toolkit and Object Relational Mapper)
-
tables 3.3.0 → 3.4.2 (Package based on HDF5 library for managing hierarchical datasets (extremely large amounts of data))
-
tblib 1.3.0 → 1.3.2 (Traceback serialization library.)
-
tensorflow 1.0.1 → 1.1.0 (TensorFlow helps the tensors flow)
-
tornado 4.4.2 → 4.5.1 (Scalable, non-blocking web server and tools (required for IPython notebook))
-
webencodings 0.5 → 0.5.1 (Character encoding aliases for legacy web content)
-
winpython 1.7.20170128 → 1.9.20170423 (WinPython distribution tools, including WPPM (package manager))
-
xarray 0.9.1 → 0.9.5 (N-D labeled arrays and datasets in Python)
-
zict 0.1.1 → 0.1.2 (Mutable mapping tools)
Removed packages:
- nose 1.3.7 (nose is a discovery-based unittest extension (e.g. NumPy test module is using nose))
I have been testing WinPython-64bit-3.5.3.2Qt5b2, with matplotlib upgraded to v2.0.2. Spyder crashed while using the editor. If I figure out how to reproduce the bug, I will post a PR to the Spyder team. Otherwise, build 2 is working fine for me.
Yes:
- as much as possible, I try to get the same thing in all distros, so the switch to the newer WinPython is painless,
- cgohlke decided Tensorflow was compatible with Python-3.6,
- it's a bit lke the rear-camera option on cars: you can't affford not to have it,
- Theano/pymc3 shame on Windows-Python-3.5/3.6, because of gcc blocage per Microsoft legal wing, leaves no choice but to look at greener pastures, Tensorflow/Edward was the one I foresee, maybe you have a better idea ?.
10 remaining items
rc (ffmpeg) : 2018-08-12/13
main change since last betas:
- going back to ipywidgets-6.0: ipywidgets-7.0 and its ecosystem are weeks from ready,
- notebook-5.1rc, spyder-3.1.2dev20170813am,
- qtpy-1.3
- don't upgrade to pywinpty-0.1.4, as it highlights issues on pywin32/pypiwin32 twin wheels (two workarounds to the same initial "pip install pywin32")
- ffmpeg.exe
main change since last beta:
- going back to ipywidgets-6.0: ipywidgets-7.0 and its ecosystem are weeks from ready,
- notebook-5.1rc, spyder-3.1.2dev20170813am,
- qtpy-1.3
- ffmpeg.exe
Interesting part to check:
- you may need to clear your browser cache with jupyter-notebook-5.1rc
- overall stability, latest python-3.5.4,
- examples of matplotlib using with ffmpeg,
- Jupyterlab infrastructure (launcher, nodejs), in anticipation of jupyterlab-1.0: errors/missing details are very possible.
MD5 SHA-1 SHA-256 Binary Size 57f519f3a59d6f9a522d7895f6d93b73 63ee0e28c55d6dfa2f8b23d243d9dad316ec7d31 a2686620242d28a63e5a8b19bc2bb8e23580a74342b41b9f9513892995ffa39d WinPython-64bit-3.5.4.0Zerorc.exe 24 639 906 Bytes 301209b81c1833e7fe22400e6e8ce23d af1e6aa6a8815d927fa23d7b9585cf97dfba2264 f366804990941728668aef6fb3ab447ee9fcf2f32adbcdac726fdf04a2672cbd WinPython-32bit-3.5.4.0Zerorc.exe 23 868 864 Bytes 53d9554d72c057e30cc3b4aecddd6dd2 7d46c27cd856d76b97d6bf0378278571cd78f76d 784512b39bf9372f97c915ef963517730940f4987731d103863e1a662cc0286e WinPython-64bit-3.6.2.0Zerorc.exe 25 312 688 Bytes bd17685f070f481a40dbb0a710ca5ae7 a8a95caef67cb4862250000e084c75b252e54165 24b4c93d34e61bda891a3ff0c078a5541546e20d7959d4c8a195e6aa05a44c79 WinPython-32bit-3.6.2.0Zerorc.exe 24 567 330 Bytes a84c043dc03494508cc94d24fc4e3307 1dde84ae43165e399fb89e6aba58055dbaf00ec2 9d0a4ff0d3e954e00f56d678bfdebc68f69f670af9a63ba45766ddbd56d64fbf WinPython-32bit-3.5.4.0Qt5rc.exe 319 995 571 Bytes 601da21869bf4b96d7335df6af606952 bf7ea12ca43d1aaf433dbefc9d081239ffbec781 36fe03bb74e16d84af0632bb12dffd54dfb7963bd9431f81d09b828caa0f64aa WinPython-64bit-3.5.4.0Qt5rc.exe 384 171 045 Bytes 1576024f6d6e6b811f990ce15da1ec0f d9f3f77d3849ec3b2d94be715c4116622f1b275c 2800417b569785b5566d2be08f1b39a0d6458a42544e0ba1f2090b57a2338bd6 WinPython-32bit-3.6.2.0Qt5rc.exe 321 304 310 Bytes b28b03d28c0ef6b543de40903190cce9 93b7819995cfc57038c6b9826a047bebd6d0a3d5 e916f435a14bee4d8763e42c5d2aa1aecc78ac7c731307e15229f7c1abd260a3 WinPython-64bit-3.6.2.0Qt5rc.exe 379 864 857 Bytes History of changes for WinPython-64bit 3.5.4.0Qt5
The following changes were made to WinPython-64bit distribution since version 3.5.3.1Qt5.
Tools
New packages:
-
ffmpeg 3.2.4 (a collection of libraries and tools to process multimedia content such as audio, video, subtitles and related metadata)
-
Nodejs v6.11.1 (a JavaScript runtime built on Chrome's V8 JavaScript engine)
Python packages
New packages:
-
backports.weakref 1.0rc1 (Backport of new features in Python's weakref module)
-
bkcharts 0.2 (High level chart types built on top of Bokeh)
-
brotli 0.6.0 (Python binding of the Brotli compression library)
-
colorcet 0.9.1 (A set of useful perceptually uniform colormaps for plotting scientific data)
-
coloredlogs 7.3 (Colored terminal output for Python's logging module)
-
cvxcanon 0.1.1 (common operations for convex optimization modeling tools.)
-
cvxpy 0.4.10 (A domain-specific language for modeling convex optimization problems in Python)
-
datashader 0.5.0 (Data visualization toolchain based on aggregating into a grid)
-
dill 0.2.7.1 (serialize all of python (almost))
-
ecos 2.0.4 (This is the Python package for ECOS: Embedded Cone Solver)
-
enum34 1.1.6 (Python 3.4 Enum backported to 3.3, 3.2, 3.1, 2.7, 2.6, 2.5, and 2.4)
-
fast_histogram 0.2.1 (Fast 1D and 2D histogram functions in Python)
-
fastcache 1.0.2 (C implementation of Python 3 functools.lru_cache)
-
fastparquet 0.1.1 (Python support for Parquet file format)
-
feather_format 0.3.1 (Python interface to the Apache Arrow-based Feather File Format)
-
gitdb2 2.0.0 (Git Object Database)
-
gitpython 2.1.3 (Python Git Library)
-
humanfriendly 4.4.1 (Human friendly output for text interfaces using Python)
-
idna 2.5 (Internationalized Domain Names in Applications (IDNA))
-
ipympl 0.0.5 (Matplotlib Jupyter Extension)
-
jupyter_sphinx 0.1.1 (Jupyter Sphinx Extensions)
-
jupyterlab 0.26.5 (Jupyter lab environment notebook server extension)
-
jupyterlab_launcher 0.3.1 (Jupyter Launcher)
-
keras_vis 0.4.1 (Neural network visualization toolkit for keras)
-
mizani 0.3.2 (Scales for Python)
-
moviepy 0.2.3.2 (Video editing with Python)
-
mpl_scatter_density 0.2 (Fast scatter density plots for Matplotlib)
-
mpldatacursor 0.6.2 (Interactive data cursors for Matplotlib)
-
multiprocess 0.70.5 (better multiprocessing and multithreading in python)
-
mypy 0.521 (Optional static typing for Python)
-
palettable 3.0.0 (Color palettes for Python)
-
plotnine 0.2.1 (A grammar of graphics for python)
-
ply 3.10 (Python Lex & Yacc)
-
pweave 0.30.0a1 (Scientific reports with embedded python computations with reST, LaTeX or markdown)
-
pyarrow 0.5.0 (Python library for Apache Arrow)
-
pybind11 2.1.1 (Seamless operability between C++11 and Python)
-
pygame 1.9.3 (Pygame gives multimedia to python.)
-
pyreadline 2.0 (IPython needs this module to display color text in Windows command window)
-
python_snappy 0.5.1 (Python library for the snappy compression library from Google)
-
pywinpty 0.1.3 (Python bindings for the winpty library)
-
pywinusb 0.4.2 (USB / HID windows helper library)
-
rope 0.10.5 (a python refactoring library...)
-
scs 1.2.6 (scs: splitting conic solver)
-
smmap2 2.0.3 (A pure python implementation of a sliding window memory map manager)
-
sphinxcontrib_websupport 1.0.1 (Sphinx API for Web Apps)
-
spyder_notebook 0.2.0.dev0 (Jupyter notebook integration with Spyder)
-
spyder_reports 0.1.0.dev0 (pyder plugin to render Markdown reports using Pweave as a backend)
-
spyder_terminal 0.2.0 (Spyder Plugin for displaying a virtual terminal (OS independent) inside the main Spyder window)
-
supersmoother 0.4 (Python implementation of Friedman's Supersmoother)
-
thriftpy 0.3.9 (Pure python implementation of Apache Thrift.)
-
typed_ast 1.0.4 (a fork of Python 2 and 3 ast modules with type comment support)
-
urllib3 1.21.1 (HTTP library with thread-safe connection pooling, file post, and more.)
-
vitables 3.0.0 (Graphical tool for browsing and editing files in both HDF5 and PyTables formats)
Upgraded packages:
-
algopy 0.5.3 → 0.5.5 (Taylor Arithmetic Computation and Algorithmic Differentiation)
-
altair_widgets 0.1.1 → 0.1.2 (Altair Widgets: An interactive visualization for statistical data for Python.)
-
astroid 1.4.9 → 1.5.3 (Rebuild a new abstract syntax tree from Python's ast (required for pylint))
-
beautifulsoup4 4.5.3 → 4.6.0 (Screen-scraping library)
-
blosc 1.5.0 → 1.5.1 (Blosc data compressor)
-
bokeh 0.12.4 → 0.12.6 (Statistical and novel interactive HTML plots for Python)
-
bottleneck 1.2.0 → 1.2.1 (Fast NumPy array functions written in Cython)
-
certifi 2017.1.23 → 2017.7.27.1 (Python package for providing Mozilla's CA Bundle.)
-
chardet 2.3.0 → 3.0.4 (Universal encoding detector for Python 2 and 3)
-
cloudpickle 0.2.2 → 0.3.1 (Extended pickling support for Python objects)
-
colorama 0.3.7 → 0.3.9 (Cross-platform colored terminal text)
-
cx_freeze 5.0.1 → 5.0.2 (Deployment tool which converts Python scripts into stand-alone Windows executables (i.e. target machine does not require Python or any other library to be installed))
-
cython 0.25.2 → 0.26 (Cython is a language that makes writing C extensions for the Python language as easy as Python)
-
dask 0.14.1 → 0.15.1 (Minimal task scheduling abstraction)
-
db.py 0.5.2 → 0.5.3 (a db package that doesn't suck)
-
decorator 4.0.11 → 4.1.2 (Better living through Python with decorators)
-
distributed 1.16.1 → 1.18.0 (Distributed computing)
-
docutils 0.13.1 → 0.14 (Text processing system for processing plaintext documentation into useful formats, such as HTML or LaTeX (includes reStructuredText))
-
edward 1.2.4 → 1.3.3 (A library for probabilistic modeling, inference, and criticism. Deep generative models, variational inference. Runs on TensorFlow.)
-
entrypoints 0.2.2 → 0.2.3 (Discover and load entry points from installed packages)
-
flask 0.12 → 0.12.2 (A microframework based on Werkzeug, Jinja2 and good intentions)
-
flask_cors 2.1.2 → 3.0.3 (A Flask extension adding a decorator for CORS support)
-
holoviews 1.7.dev8 → 1.8.2 (Composable, declarative data structures for building complex visualizations easily.)
-
imageio 2.1.2 → 2.2.0 (Library for reading and writing a wide range of image, video, scientific, and volumetric data formats.)
-
ipykernel 4.5.2 → 4.6.1 (IPython Kernel for Jupyter)
-
ipython 5.3.0 → 6.1.0 (Enhanced Python shell)
-
isort 4.2.5 → 4.2.15 (A Python utility / library to sort Python imports.)
-
jedi 0.10.0 → 0.10.2 (An autocompletion tool for Python that can be used for text editors)
-
jinja2 2.9.5 → 2.9.6 (Sandboxed template engine (provides a Django-like non-XML syntax and compiles templates into executable python code))
-
jmespath 0.9.2 → 0.9.3 (JSON Matching Expressions)
-
julia 0.1.1 → 0.1.2 (Python interface to the Julia language)
-
jupyter_client 5.0.0 → 5.1.0 (Jupyter protocol implementation and client libraries)
-
keras 2.0.2 → 2.0.6 (Theano-based Deep Learning library)
-
lazy_object_proxy 1.2.2 → 1.3.1 (A fast and thorough lazy object proxy.)
-
llvmlite 0.16.0 → 0.19.0 (lightweight wrapper around basic LLVM functionality)
-
lmfit 0.9.5 → 0.9.7 (Least-Squares Minimization with Bounds and Constraints)
-
lxml 3.7.3 → 3.8.0 (Powerful and Pythonic XML processing library combining libxml2/libxslt with the ElementTree API.)
-
matplotlib 2.0.0 → 2.0.2 (2D plotting library (embeddable in GUIs created with PyQt))
-
metakernel 0.20.1 → 0.20.4 (Metakernel for Jupyter)
-
nbconvert 5.1.1 → 5.2.1 (Converting Jupyter Notebooks)
-
nbdime 0.2.0 → 0.3.0 (Tools for diffing and merging of Jupyter notebooks)
-
netcdf4 1.2.7 → 1.2.9 (python/numpy interface to netCDF library (versions 3 and 4))
-
nltk 3.2.2 → 3.2.4 (The Natural Language Toolkit (NLTK) is a Python package for natural language processing.)
-
notebook 5.0.0rc2 → 5.1.0rc1 (# Jupyter Notebook)
-
numba 0.31.0 → 0.34.0 (compiling Python code using LLVM)
-
numpy 1.11.3+mkl → 1.13.1+mkl (NumPy: multidimensional array processing for numbers, strings, records and objects (SciPy''s core module))
-
numpydoc 0.6.0 → 0.7.0 (Sphinx extension to support docstrings in Numpy format)
-
oct2py 4.0.5 → 4.0.6 (Python to GNU Octave bridge --> run m-files from python.)
-
pandas 0.19.2 → 0.20.3 (Powerful data structures for data analysis, time series and statistics)
-
pandas_datareader 0.3.0.post0 → 0.5.0 (Data readers extracted from the pandas codebase,should be compatible with recent pandas versions)
-
pandocfilters 1.4.1 → 1.4.2 (Utilities for writing pandoc filters in python)
-
param 1.5.0 → 1.5.1 (Declarative Python programming using Parameters.)
-
partd 0.3.7 → 0.3.8 (Appendable key-value storage)
-
pillow 4.0.0 → 4.2.1 (Python Imaging Library (fork))
-
prompt_toolkit 1.0.13 → 1.0.15 (Library for building powerful interactive command lines in Python)
-
protobuf 3.2.0 → 3.3.0 (Protocol Buffers - Google's data interchange format)
-
psutil 5.2.0 → 5.2.2 (Provides an interface for retrieving information on all running processes and system utilization (CPU, disk, memory, network) in a portable way)
-
ptpython 0.36 → 0.41 (Python REPL build on top of prompt_toolkit)
-
ptyprocess 0.5.1 → 0.5.2 (Run a subprocess in a pseudo terminal)
-
pulp 1.6.5 → 1.6.8 (PuLP is an LP modeler written in python. PuLP can generate MPS or LP files and call GLPK, COIN CLP/CBC, CPLEX, and GUROBI to solve linear problems)
-
py 1.4.33 → 1.4.34 (library with cross-python path, ini-parsing, io, code, log facilities)
-
pyflakes 1.5.0 → 1.6.0 (passive checker of Python programs)
-
pyflux 0.4.14 → 0.4.15 (Open source time series library for Python)
-
pylint 1.6.5 → 1.7.2 (Logilab code analysis module: analyzes Python source code looking for bugs and signs of poor quality)
-
pymc3 3.0 → 3.1 (Markov Chain Monte Carlo sampling toolkit.)
-
pymongo 3.4.0 → 3.5.0 (Python driver for MongoDB http://www.mongodb.org)
-
pyodbc 4.0.15 → 4.0.17 (DB API Module for ODBC)
-
pyqt5 5.8.1 → 5.9 (Python bindings for the Qt cross platform GUI toolkit)
-
pyserial 3.3 → 3.4 (Library encapsulating the access for the serial port)
-
pytest 3.0.7 → 3.2.0 (pytest: simple powerful testing with Python)
-
Python 3.5.3 → 3.5.4 (Python programming language with standard library)
-
python_dateutil 2.6.0 → 2.6.1 (Powerful extensions to the standard datetime module)
-
pytz 2016.10 → 2017.2 (World Timezone Definitions for Python)
-
pywin32 220.1 → 221 (Python library for Windows)
-
qtconsole 4.2.1 → 4.3.0 (Jupyter Qt console)
-
qtpy 1.2.1 → 1.3.0 (Provides an abstraction layer on top of the various Qt bindings (PyQt5, PyQt4 and PySide) and additional custom QWidgets.)
-
requests 2.13.0 → 2.18.3 (Requests is an Apache2 Licensed HTTP library, written in Python, for human beings.)
-
requests_file 1.4.1 → 1.4.2 (File transport adapter for Requests)
-
requests_toolbelt 0.7.1 → 0.8.0 (A utility belt for advanced users of python-requests)
-
rpy2 2.8.5 → 2.8.6 (Python interface to the R language (embedded R))
-
rx 1.5.8 → 1.5.9 (Reactive Extensions (Rx) for Python)
-
s3fs 0.0.9 → 0.1.2 (Convenient Filesystem interface over S3)
-
scikit_image 0.13.dev0 → 0.13.0 (Image processing toolbox for SciPy)
-
scikit_learn 0.18.1 → 0.18.2 (A set of Python modules for machine learning and data mining)
-
scipy 0.19.0 → 0.19.1 (SciPy: Scientific Library for Python (advanced math, signal processing, optimization, statistics, ...))
-
seaborn 0.8.dev0 → 0.8.0 (statistical data visualization)
-
setuptools 34.3.2 → 36.2.7 (Download, build, install, upgrade, and uninstall Python packages - easily)
-
simplejson 3.10.0 → 3.11.1 (Simple, fast, extensible JSON (JavaScript Object Notation) encoder/decoder)
-
sip 4.19.1 → 4.19.3 (Python extension module generator for C and C++ libraries)
-
sounddevice 0.3.7 → 0.3.8 (Play and Record Sound with Python)
-
sphinx 1.5.2 → 1.6.3 (Tool for generating documentation which uses reStructuredText as its markup language)
-
spyder 3.1.4.dev0 → 3.2.1.dev0 (Scientific PYthon Development EnviRonment: designed for interactive computing and data visualisation with a simple and intuitive user interface)
-
sqlalchemy 1.1.6 → 1.1.13 (SQL Toolkit and Object Relational Mapper)
-
sympy 1.0 → 1.1.1 (Symbolic Mathematics Library)
-
tables 3.3.0 → 3.4.2 (Package based on HDF5 library for managing hierarchical datasets (extremely large amounts of data))
-
tblib 1.3.0 → 1.3.2 (Traceback serialization library.)
-
tensorflow 1.0.1 → 1.2.1 (TensorFlow helps the tensors flow)
-
testpath 0.3 → 0.3.1 (Test utilities for code working with files and commands)
-
tornado 4.4.2 → 4.5.1 (Scalable, non-blocking web server and tools (required for IPython notebook))
-
tqdm 4.11.2 → 4.15.0 (A Simple Python Progress Meter)
-
twine 1.8.1 → 1.9.1 (Collection of utilities for interacting with PyPI)
-
vega 0.4.4 → 0.5.0 (An IPython/ Jupyter widget for Vega and Vega-Lite)
-
webencodings 0.5 → 0.5.1 (Character encoding aliases for legacy web content)
-
werkzeug 0.12.1 → 0.12.2 (The Swiss Army knife of Python web development)
-
winpython 1.7.20170128 → 1.9.20170812 (WinPython distribution tools, including WPPM (package manager))
-
wordcloud 1.3.1 → 1.3.2 (A little word cloud generator)
-
xarray 0.9.1 → 0.9.6 (N-D labeled arrays and datasets in Python)
-
xlsxwriter 0.9.5 → 0.9.8 (A Python module for creating Excel XLSX files.)
-
xlwings 0.10.4 → 0.11.4 (Interact with Excel from Python and vice versa)
-
zict 0.1.1 → 0.1.2 (Mutable mapping tools)
Removed packages:
-
appdirs 1.4.3 (A small Python module for determining appropriate " + "platform-specific dirs, e.g. a "user data dir".)
-
args 0.1.0 (Command Arguments for Humans.)
-
clint 0.5.1 (Python Command Line Interface Tools)
-
commonmark 0.5.4 (Python parser for the CommonMark Markdown spec)
-
ggplot 0.11.5 (ggplot for python)
-
nbsphinx 0.2.13 (Jupyter Notebook Tools for Sphinx)
-
nose 1.3.7 (nose is a discovery-based unittest extension (e.g. NumPy test module is using nose))
-
packaging 16.8 (Core utilities for Python packages)
-
recommonmark 0.4.0 (A markdown parser for docutils)
-
rope_py3k 0.9.4.post1 (a python refactoring library...)
-
sortedcollections 0.4.2 (Python Sorted Collections)
@stonebig , How come you take development builds of Spyder?
procedure:
- go there : https://github.com/spyder-ide/spyder/tree/3.x
- click on green download button, place result in d:\toto, it will download a "d:\toto\spyder-3.x.zip"
- unzip the result in d:\toto\spyder-3.x\spyder-3.x,
- launch an existing "WinPython Command Prompt.exe"
- cd/D d;\toto\spyder-3.x\spyder-3.x
- python setup.py sdist bdist_wheel
- then you have a wheel at d:\toto\spyder-3.x\spyder-3.x\dist
- and now you can install that wheel the usual way.
also, this may be simplified, if you don't bother the wheel:
- launch an existing "WinPython Command Prompt.exe"
- pip uninstall spyder
- pip install https://github.com/spyder-ide/spyder/archive/3.x.zip
there may be simpler steps.
for pre-releases of bokeh, that doesn't work and the method to use is:
- pip install --pre -i https://pypi.anaconda.org/bokeh/channel/dev/simple bokeh
I didn't mean how you do it.
I asked why you use in your distribution the development builds instead of the official builds?The Spyder stable build is perfect for Anaconda stable-release, but sometimes lacks some bug-fixes for WinPython.
So I use the dev build of Spyder stable branch 3.x, when it contains bug-fixes I want.
I see.
Seems reasonable.They have a great progress there with Spyder, really great team.
pytorch is certainly great, trading blows with tensorflow for the benefit of Python stack. Yet I probably can't ship "CuDNN 7 + CUDA 8 / 9" with it.... But I see on the front page a "Facebook and Microsoft introduce ONNX"
Let me know when it comes to Windows, and I'll have a check
I'm not sure how it works but I think it has CuDNN built in.
So all needed is to use the one with CuDNN 7.I'm not sure about CUDA (Does it use the system library or its own).
I will surely let you know once it is available.
Thank You.
There is also the japan "Cupy/chainer" outsider that Travis Oliphant (numpy father) noticed https://twitter.com/teoliphant/status/941758883554619392
I think Tensorflow/cuda have the marketshare for this round, but things evolve quickly and cuda must loose dominance, as it's vendor-tied.
@stonebig , As you can see in ContinuumIO/anaconda-issues#7410 PyTorch is a package by itself. Coming with CuDNN + CUDA on its own.
It is a native Python Deep Learning package.
It doesn't have to be Tensorflow or PyTorch.By the way, it seems Windows support will be official in version 0.4 of PyTorch.
Good news: having all leading DeepLearning packages under Python makes Python sucessfull.
Unfortunately, Tensorflow is already getting too big for WinPython 2go limit, and Pytorch shall be bigger.
release estimate: mid-august
expected:
new packages:
package updates:
infrastructure:
WinPython-201701 final is out, and is identical to rc.