fast-neural-style
fast-neural-style copied to clipboard
jcjohnson
Unsupported HDF5 version: 1.10.0
th train.lua -h5_file ../out/h5/file.h5 -style_image images/styles/starry_night.jpg -style_image_size 384 -content_weights 1.0 -style_weights 5.0 -gpu 0 /home/kaismax/torch/install/bin/luajit: /home/kaismax/torch/install/share/lua/5.1/trepl/init.lua:389:
/home/kaismax/torch/install/share/lua/5.1/trepl/init.lua:389: /home/kaismax/torch/install/share/lua/5.1 /hdf5/ffi.lua:71: Unsupported HDF5 version: 1.10.0 i can't solve this problem.. help please ,, $luarocks list Installed rocks: .... 0-0 (installed) - /home/kaismax/torch/install/lib/luarocks/rocks ...
$ th
______ __ | Torch7 /_ / ________/ / | Scientific computing for Lua. / / / _ / __/ / _ \ | Type ? for help /_/ _// _//// | https://github.com/torch | http://torch.ch
I solved this problem by building HDF5 1.8.18 from source and installing it in /opt/hdf5-1.8.18.
wget https://support.hdfgroup.org/ftp/HDF5/releases/hdf5-1.8/hdf5-1.8.18/src/hdf5-1.8.18.tar.gz tar -xzvf hdf5-1.8.18.tar.gz cd hdf5-1.8.18 ./configure --prefix=/opt/hdf5-1.8.18 make make install
Building HDF5 prints an unholy amount of warnings from the compiler!
Then did the following:
export PATH=/opt/hdf5-1.8.18/bin:$PATH export LD_LIBRARY_PATH=/opt/hdf5-1.8.18/lib:$LD_LIBRARY_PATH luarocks remove hdf5 luarocks install hdf5 pip uninstall h5py pip install h5py
You do this because luarocks is invoking HDF5's CC wrapper script named h5cc to run the C compiler and h5cc does the script-foo to link against it's version of the libs. In the output from the "luarocks install hdf5" look for confirmation of the HDF5 location used.
I expect you'll need to export these environment variables each session you run the lua scripts.
TBH I didn't bother digging into the pip reinstall except to note that afterwards the file "/usr/local/lib/python2.7/dist-packages/h5py/version.py" contains this line "api_version_tuple = (1,8)"
train.lua is currently grinding away at a dataset!
this got me going immediately https://github.com/deepmind/torch-hdf5/issues/76#issuecomment-357379520
