But you can install it with our own channel mila-udem (that support only Python 2.7 and 3.5): conda install -c mila-udem pygpu Install the latest, development version of libgpuarray following the Step-by-step instructions.Ĭurrently, you need libgpuarray version 0.7.1 that is not in conda default channel. This is important when they have already been installed as system packages. Use no-deps when you don’t want the dependencies of Theano to be installed through pip. It will install Theano in your local site-packages. Use user for a user installation without admin rights. Install the latest, bleeding-edge, development version of Theano with: pip install git+Īny argument between is optional. Download it with: git clone Īnd then follow the Step-by-step instructions. The latest stable version of Theano is 0.9.0 (tagged with rel-0.9.0).įor the stable version of Theano you need a specific version of libgpuarray, that has been tagged v0.6.9. If you encountered any trouble, head to the Troubleshooting page. will install the requirements in order to generate the documentation. will install the requirements for testing.
Install the latest stable version of Theano with: pip install TheanoĪny argument between is optional. If you use pip, you have to install Theano and libgpuarray separately. Latest conda packages for theano ( >= 0.9) and pygpu ( >= 0.6*) currently don’t support Python 3.4 branch. Libgpuarray will be automatically installed as a dependency of pygpu. If you use conda, you can directly install both theano and pygpu.
You must reboot the computer after the driver installation.
Install and configure the GPU drivers (recommended)įollow this link to install the CUDA driver and the CUDA Toolkit. Package parameterized is also optional but may be required for unit testing. git package installs git source control through conda, which is required for the development versions of Theano and libgpuarray m2w64-toolchain package provides a fully-compatible version of GCC and is then highly recommended. Install requirements and optional packages conda install numpy scipy mkl-service libpython Īrguments between are optional.
If you want fast compiled code (recommended), make sure you have g++ installed. Requirements installation through Conda (recommended) It is faster then using an equivalent graph of Theano ops. For cuda 8, the dev version of skcuda (will be released as 0.5.2) is needed for cusolver: pip install pycuda pip install git+.
Quick install pip install pycuda scikit-cuda. Required for some extra operations on the GPU like fft and solvers. Required for GPU/CPU code generation on CUDA and OpenCL devices (see: GpuArray Backend). Highly recommended Required for GPU code generation/execution on NVIDIA gpus. LaTeX and dvipng are also necessary for math to show up as images. Theano can fall back on a NumPy-based Python execution model, but a C compiler allows for vastly faster execution.įor building the documentation. GCC compiler with g++ (version >= 4.2.*), and Python development files Alternatively, we suggest to install OpenBLAS, with the development headers ( -dev, -devel, depending on your Linux distribution). Recommended: MKL, which is free through Conda with mkl-service package. Python = 2.7* or ( >= 3.4 and = 1.9.1 = 0.14 =0.8 could work, but earlier versions have known bugs with sparse matrices.īLAS installation (with Level 3 functionality)
We only support the installation of the requirements through conda.
If you want to install the bleeding-edge or development version of Theano from GitHub, please make sure you are reading the latest version of this page.