# Copyright 2013-2021 Lawrence Livermore National Security, LLC and other # Spack Project Developers. See the top-level COPYRIGHT file for details. # # SPDX-License-Identifier: (Apache-2.0 OR MIT) import json from spack import * class PyChainer(PythonPackage): """ Chainer is a Python-based deep learning framework aiming at flexibility. It provides automatic differentiation APIs based on the define-by-run approach (a.k.a. dynamic computational graphs) as well as object-oriented high-level APIs to build and train neural networks. It also supports CUDA/cuDNN using CuPy for high performance training and inference. """ homepage = "https://chainer.org/" url = "https://github.com/chainer/chainer/archive/v7.2.0.tar.gz" maintainers = ['adamjstewart'] version('7.2.0', sha256='6e2fba648cc5b8a5421e494385b76fe5ec154f1028a1c5908557f5d16c04f0b3') version('6.7.0', sha256='87cb3378a35e7c5c695028ec91d58dc062356bc91412384ea939d71374610389') variant("mn", default=False, description="run with ChainerMN") depends_on('python@3.5.1:', when='@7:', type=('build', 'run')) depends_on('py-setuptools', type=('build', 'run')) depends_on('py-numpy@1.9:', type=('build', 'run')) depends_on('py-six@1.9.0:', type=('build', 'run')) depends_on('py-typing-extensions', type=('build', 'run')) depends_on('py-typing-extensions@:3.6.6', when='@:6', type=('build', 'run')) depends_on('py-filelock', type=('build', 'run')) depends_on('py-protobuf@3:', type=('build', 'run')) depends_on('py-typing@:3.6.6', when='@:6', type=('build', 'run')) # Dependencies only required for test of ChainerMN depends_on('py-matplotlib', type=('build', 'run'), when='+mn') depends_on('py-mpi4py', type=('build', 'run'), when='+mn') depends_on("mpi", type=("build", "run"), when='+mn') @run_after('install') def cache_test_sources(self): if '+mn' in self.spec: self.cache_extra_test_sources("examples") def test(self): if "+mn" in self.spec: # Run test of ChainerMN test_dir = self.test_suite.current_test_data_dir mnist_dir = join_path( self.install_test_root, "examples", "chainermn", "mnist" ) mnist_file = join_path(mnist_dir, "train_mnist.py") mpi_name = self.spec["mpi"].prefix.bin.mpirun python_exe = self.spec["python"].command.path opts = [ "-n", "4", python_exe, mnist_file, "-o", test_dir, ] env["OMP_NUM_THREADS"] = "4" self.run_test( mpi_name, options=opts, work_dir=test_dir, ) # check results json_open = open(join_path(test_dir, 'log'), 'r') json_load = json.load(json_open) v = dict([(d.get('epoch'), d.get('main/accuracy')) for d in json_load]) if 1 not in v or 20 not in v: raise RuntimeError('Cannot find epoch 1 or epoch 20') if abs(1.0 - v[1]) < abs(1.0 - v[20]): raise RuntimeError('ChainerMN Test Failed !')