Unsupported HDF5 version: 1.10.0

[jnhwkim]

It seems torch-hdf5 doesn't support 1.10.0 now.

$ th -lhdf5
could not load hdf5, skipping
/usr/local/torch/install/share/lua/5.1/hdf5/ffi.lua:71: Unsupported HDF5 version: 1.10.0
stack traceback:
    [C]: in function 'error'
    /usr/local/torch/install/share/lua/5.1/hdf5/ffi.lua:71: in function 'checkHDF5Version'
    /usr/local/torch/install/share/lua/5.1/hdf5/ffi.lua:78: in main chunk
    [C]: in function 'dofile'
    /usr/local/torch/install/share/lua/5.1/torch/init.lua:54: in function 'include'
    /usr/local/torch/install/share/lua/5.1/hdf5/init.lua:29: in main chunk
    [C]: in function 'require'
    ...ocal/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:48: in function <...ocal/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:48>
    [C]: in function 'xpcall'
    ...ocal/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:48: in main chunk
    [C]: at 0x00406670

[TimZaman]

Duplicate. This issue was already raised in #67 and closed because they have no plans to support it. Not a valid reason to close the issue imo. I'll try to get a PR for this because this 1.10->1.8 thing is a real pita for me. Am I correct in thinking deepmind's movement to TensorFlow means the torch packages and deepmind github presence will not develop?

[rave78]

I found the following solution https://www.bountysource.com/issues/34861873-unsupported-hdf5-version. I have not tried yet.

[TimZaman]

That page does not yield a solution to the actual compatibility with a more recent version of the hdf5 library. That page just tells you to install an old version.

[cdluminate]

such a pity. Debian has transited to HDF5-1.10.

[hassanshallal]

This config.lua solved the issue in my system:

hdf5._config = {
HDF5_INCLUDE_PATH = "/Users/yourname/anaconda/pkgs/hdf5-1.8.17-1/include",
HDF5_LIBRARIES = "/Users/yourname/anaconda/pkgs/hdf5-1.8.17-1/lib/libhdf5.dylib;/usr/lib/libpthread.dylib;/usr/lib/libz.dylib;/usr/lib/libdl.dylib;/usr/lib/libm.dylib"
}

[FrenchBully]

Is there still no solution for this issue?

[hassanshallal]

Do you have anaconda installed?

…

On Apr 6, 2017, at 4:35 PM, Kevin ***@***.***> wrote: Is there still no solution for this issue? — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#76 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ASUu5JZZWCdeexv8QqnYvpdv3pQeV2Kwks5rtXaogaJpZM4Jeg2E>.

[FrenchBully]

Actually was able to get it working thanks, but with an older version of hdf5. Credit goes to Tom Schofield,
Steps:

1. brew install hdf5@1.8
2. copy contents from usr/local/Cellar/hdf5@1.8 over to usr/local/Cellar/hdf5
3. brew switch hdf5 1.8.18
4. Set /usr/local/torch/install/share/lua/5.1/hdf5/config.lua
   hdf5._config = { HDF5_INCLUDE_PATH = "/usr/local/include", HDF5_LIBRARIES = "/usr/local/lib/libhdf5.dylib;/usr/local/lib/libsz.dylib;/usr/lib/libz.dylib;/usr/lib/libdl.dylib;/usr/lib/libm.dylib" }

[Cadene]

I solved the issue by installing a certain version of hdf5, but not the 1.10 version of hdf5.

1. Install anaconda
2. Create a new env : conda create -n hdf5 python=3.6
3. Install hdf5 : conda install hdf5=1.8.17
4. Modify torch-hdf5 config file : nano ~/torch/install/share/lua/5.1/hdf5/config.lua
   (I just added the HDF5_INCLUDE_PATH and modified the path to libhdf5.so in HDF6_LIBRARIES)

hdf5._config = {
    HDF5_INCLUDE_PATH = "/home/cadene/anaconda3/envs/hdf5/include/",
    HDF5_LIBRARIES = "/home/cadene/anaconda3/envs/hdf5/lib/libhdf5.so;/usr/lib/x86_64-linux-gnu/librt.so;/usr/lib/x86_64-linux-gnu/libpthread.so;/usr/lib/x86_64-linux-gnu/libz.so;/usr/lib/x86_64-linux-gnu/libdl.so;/usr/lib/x86_64-linux-gnu/libm.so"
}

5. Try : th -e "require 'hdf5'"
6. Dance if no error or cry and try harder

[MelomanCool]

There is a fork that supports 1.10: https://github.com/anibali/torch-hdf5/tree/hdf5-1.10
You can install it like so:

git clone https://github.com/anibali/torch-hdf5.git
cd torch-hdf5
git checkout hdf5-1.10 
luarocks make hdf5-0-0.rockspec

[agilebean]

Usually, installing the hdf5 bindings via luarocks is the more direct way and should work:

luarocks install https://raw.githubusercontent.com/deepmind/torch-hdf5/master/hdf5-0-0.rockspec

[Wentong-DST]

I found another way to solve this, which works for me:
First, download hdf5-1.8.18 source file
Second, use LD_LIBRARY_PATH="../hdf5-1.8.18/src" th xxx.lua when running torch scripts