Theano-MPI

Theano-MPI is a python framework for distributed training of deep learning models built in Theano. It implements data-parallelism in serveral ways, e.g., Bulk Synchronous Parallel, Elastic Averaging SGD and Gossip SGD. This project is an extension to theano_alexnet, aiming to scale up the training framework to more than 8 GPUs and across nodes. Please take a look at this technical report for an overview of implementation details. To cite our work, please use the following bibtex entry.

@article{ma2016theano,
  title = {Theano-MPI: a Theano-based Distributed Training Framework},
  author = {Ma, He and Mao, Fei and Taylor, Graham~W.},
  journal = {arXiv preprint arXiv:1605.08325},
  year = {2016}
}

Theano-MPI is compatible for training models built in different framework libraries, e.g., Lasagne, Keras, Blocks, as long as its model parameters can be exposed as theano shared variables. Theano-MPI also comes with a light-weight layer library for you to build customized models. See wiki for a quick guide on building customized neural networks based on them. Check out the examples of building Lasagne VGGNet, Wasserstein GAN, LS-GAN and Keras Wide-ResNet.

Dependencies

Theano-MPI depends on the following libraries and packages. We provide some guidance to the installing them in wiki.

• OpenMPI 1.8 + or an MPI-2 standard equivalent that supports CUDA.
• mpi4py built on OpenMPI.
• numpy
• Theano 0.9 +
• zeromq
• hickle
• CUDA 7.5 +
• cuDNN a version compatible with your CUDA Installation.
• pygpu
• NCCL

Installation

Once all dependeices are ready, one can clone Theano-MPI and install it by the following.

 $ python setup.py install [--user]

Usage

To accelerate the training of Theano models in a distributed way, Theano-MPI tries to identify two components:

• the iterative update function of the Theano model
• the parameter sharing rule between instances of the Theano model

It is recommended to organize your model and data definition in the following way.

• launch_session.py or launch_session.cfg
• models/*.py
  • __init__.py
  • modelfile.py : defines your customized ModelClass
  • data/*.py
    • dataname.py : defines your customized DataClass

Your ModelClass in modelfile.py should at least have the following attributes and methods:

• self.params : a list of Theano shared variables, i.e. trainable model parameters
• self.data : an instance of your customized DataClass defined in dataname.py
• self.compile_iter_fns : a method, your way of compiling train_iter_fn and val_iter_fn
• self.train_iter : a method, your way of using your train_iter_fn
• self.val_iter : a method, your way of using your val_iter_fn
• self.adjust_hyperp : a method, your way of adjusting hyperparameters, e.g., learning rate.
• self.cleanup : a method, necessary model and data clean-up steps.

Your DataClass in dataname.py should at least have the following attributes:

• self.n_batch_train : an integer, the amount of training batches needed to go through in an epoch
• self.n_batch_val : an integer, the amount of validation batches needed to go through during validation

After your model definition is complete, you can choose the desired way of sharing parameters among model instances:

• BSP (Bulk Syncrhonous Parallel)
• EASGD (Elastic Averaging SGD)
• GOSGD (Gossip SGD)

Below is an example launch config file for training a customized ModelClass on two GPUs.

# launch_session.cfg
RULE=BSP
MODELFILE=models.modelfile
MODELCLASS=ModelClass
DEVICES=cuda0,cuda1

Then you can launch the training session by calling the following command:

 $ tmlauncher -cfg=launch_session.cfg

Alternatively, you can launch sessions within python as shown below:

# launch_session.py
from theanompi import BSP

rule=BSP()
# modelfile: the relative path to the model file
# modelclass: the class name of the model to be imported from that file
rule.init(devices=['cuda0', 'cuda1'] , 
          modelfile = 'models.modelfile', 
          modelclass = 'ModelClass') 
rule.wait()

More examples can be found here.

Example Performance

BSP tested on up to eight Tesla K80 GPUs

Training (+communication) time per 5120 images in seconds: [allow_gc = True, using nccl32 on copper]

More details on the benchmark can be found in this notebook.

Note

• Test your single GPU model with theanompi/models/test_model.py before trying data-paralle rule.

• You may want to use those helper functions in /theanompi/lib/opt.py to construct optimizers in order to avoid common pitfalls mentioned in (#22) and get better convergence.

• Binding cores according to your NUMA topology may give better performance. Try the -bind option with the launcher (needs hwloc depedency).

• Using the launcher script is prefered to start training. Using python to start training currently cause core binding problem especially on a NUMA system.

• Shuffling training examples before asynchronous training makes the loss surface a lot smoother during model converging.

• Some known bugs and possible enhancement are listed in Issues. We welcome all kinds of participation (bug reporting, discussion, pull request, etc) in improving the framework.

License

© Contributors, 2016-2017. Licensed under an ECL-2.0 license.