Merge branch 'master' into feat/fixtures

Author: Vasileios Karakasis

cscs-checks/apps/jupyter/check_ipcmagic.py

@@ -4,86 +4,26 @@
 # SPDX-License-Identifier: BSD-3-Clause
 
 import reframe as rfm
-import reframe.utility.osext as osext
-import reframe.utility.sanity as sn
-from reframe.core.backends import getlauncher
+
+from hpctestlib.apps.jupyter.ipcmagic import ipcmagic_check
 
 
 @rfm.simple_test
-class IPCMagicCheck(rfm.RunOnlyRegressionTest):
-    descr = 'Distributed training with TensorFlow using ipyparallel'
+class cscs_ipcmagic_check(ipcmagic_check):
     valid_systems = ['daint:gpu', 'dom:gpu']
-    valid_prog_environs = ['PrgEnv-gnu']
-    modules = [
-        f'ipcmagic', f'jupyterlab',
-        f'Horovod/0.21.0-CrayGNU-{osext.cray_cdt_version()}-tf-2.4.0'
-    ]
-    num_tasks = 2
-    num_tasks_per_node = 1
-    executable = 'ipython'
-    executable_opts = ['tf-hvd-sgd-ipc-tf2.py']
+    valid_prog_environs = ['builtin']
+    modules = ['jupyterlab', 'Horovod']
+    maintainers = ['RS', 'TR']
+    tags = {'production'}
     reference = {
         'daint:gpu': {
             'slope': (2.0, -0.1, 0.1, 'N/A'),
             'offset': (0.0, -0.1, 0.1, 'N/A'),
             'retries': (0, None, None, 'N/A'),
-            'time': (10, None, None, 's'),
         },
         'dom:gpu': {
             'slope': (2.0, -0.1, 0.1, 'N/A'),
             'offset': (0.0, -0.1, 0.1, 'N/A'),
             'retries': (0, None, None, 'N/A'),
-            'time': (10, None, None, 's'),
         }
     }
-
-    maintainers = ['RS', 'TR']
-    tags = {'production'}
-
-    @run_after('setup')
-    def daint_module_workaround(self):
-        if self.current_system.name == 'daint':
-            # FIXME: Use the default modules once Dom/Daint are aligned
-            self.modules = [
-                f'ipcmagic/1.0.1-CrayGNU-{osext.cray_cdt_version()}',
-                f'Horovod/0.21.0-CrayGNU-{osext.cray_cdt_version()}-tf-2.4.0'
-            ]
-            # FIXME: Enforce loading of jupyterlab module since
-            # `module show jupyterlab` throws a Tcl error on Daint
-            self.prerun_cmds = ['module load jupyterlab']
-
-    @sanity_function
-    def assert_successful_execution(self):
-        nids = sn.extractall(r'nid(?P<nid>\d+)', self.stdout, 'nid', str)
-        return sn.all([
-            sn.assert_ne(nids, []), sn.assert_ne(nids[0], nids[1]),
-            sn.assert_found(r'IPCluster is ready\!\s+', self.stdout),
-            sn.assert_found(r'slope=\S+', self.stdout)
-        ])
-
-    @performance_function('N/A')
-    def slope(self):
-        return sn.extractsingle(r'slope=(?P<slope>\S+)', self.stdout,
-                                'slope', float)
-
-    @performance_function('N/A')
-    def offset(self):
-        return sn.extractsingle(r'offset=(?P<offset>\S+)', self.stdout,
-                                'offset', float)
-
-    @performance_function('N/A')
-    def retries(self):
-        return 4 - sn.count(sn.findall(r'IPCluster is already running',
-                                       self.stdout))
-
-    @performance_function('s')
-    def time(self):
-        return sn.extractsingle(r'IPCluster is ready\!\s+'
-                                r'\((?P<time>\d+) seconds\)',
-                                self.stdout, 'time', float)
-
-    @run_before('run')
-    def reset_launcher(self):
-        # Change the job launcher since `ipython`
-        # needs to be launched without `srun`.
-        self.job.launcher = getlauncher('local')()

docs/hpctestlib.rst

@@ -1,5 +1,7 @@
-ReFrame Test Library
-====================
+ReFrame Test Library (experimental)
+===================================
+
+This is a collection of generic tests that you can either run out-of-the-box by specializing them for your system using the :option:`-S` option or create your site-specific tests by building upon them.
 
 
 Scientific Applications
@@ -16,3 +18,11 @@ Python
 .. automodule:: hpctestlib.python.numpy.numpy_ops
    :members:
    :show-inheritance:
+
+
+Interactive Computing
+-----------------------
+
+.. automodule:: hpctestlib.apps.jupyter.ipcmagic
+   :members:
+   :show-inheritance:

hpctestlib/apps/jupyter/ipcmagic/__init__.py

@@ -0,0 +1,67 @@
+# Copyright 2016-2021 Swiss National Supercomputing Centre (CSCS/ETH Zurich)
+# ReFrame Project Developers. See the top-level LICENSE file for details.
+#
+# SPDX-License-Identifier: BSD-3-Clause
+
+import reframe as rfm
+import reframe.utility.sanity as sn
+
+from reframe.core.backends import getlauncher
+
+
+@rfm.simple_test
+class ipcmagic_check(rfm.RunOnlyRegressionTest, pin_prefix=True):
+    '''Test ipcmagic via a distributed TensorFlow training with ipyparallel.
+
+    `ipcmagic <https://github.com/eth-cscs/ipcluster_magic>`__ is a Python
+    package and collection of CLI scripts for controlling clusters for
+    Jupyter. For more information, please have a look
+    `here <https://user.cscs.ch/tools/interactive/jupyterlab/#ipython>`__.
+
+    This test checks the ipcmagic performance.
+    To do this, a single-layer neural network is trained against a noisy linear
+    function. The parameters of the fitted linear function are returned in the
+    end along with the resulting loss function. The default assumption is that
+    ipcmagic is already installed on the system under test.
+
+    '''
+
+    executable = 'ipython'
+    executable_opts = ['tf-hvd-sgd-ipc-tf2.py']
+    num_tasks = 2
+    num_tasks_per_node = 1
+    descr = 'Distributed training with TensorFlow using ipyparallel'
+
+    @performance_function('N/A')
+    def fitted_line_slope(self):
+        return sn.extractsingle(r'slope=(?P<slope>\S+)',
+                                self.stdout, 'slope', float)
+
+    @performance_function('N/A')
+    def fitted_line_offset(self):
+        return sn.extractsingle(r'offset=(?P<offset>\S+)',
+                                self.stdout, 'offset', float)
+
+    @performance_function('N/A')
+    def retries(self):
+        return 4 - sn.count(sn.findall(r'IPCluster is already running',
+                                       self.stdout))
+
+    @run_before('run')
+    def reset_launcher(self):
+        # Change the job launcher since `ipython`
+        # needs to be launched without `srun`.
+        self.job.launcher = getlauncher('local')()
+
+    @sanity_function
+    def assert_successful_execution(self):
+        '''Checks that the program is running on 2 different nodes (nids
+        are different), that IPCMagic is configured and returns the correct
+        end-of-program message (returns the slope parameter in the end).'''
+
+        nids = sn.extractall(r'nid(?P<nid>\d+)', self.stdout, 'nid', str)
+        return sn.all([
+            sn.assert_eq(sn.len(nids), 2),
+            sn.assert_ne(nids[0], nids[1]),
+            sn.assert_found(r'slope=\S+', self.stdout)
+        ])

cscs-checks/apps/jupyter/src/tf-hvd-sgd-ipc-tf2.py renamed to hpctestlib/apps/jupyter/ipcmagic/src/tf-hvd-sgd-ipc-tf2.py

@@ -1420,7 +1420,8 @@ def _clone_to_stagedir(self, url):
         self.logger.debug(f'Cloning URL {url} into stage directory')
         osext.git_clone(
             self.sourcesdir, self._stagedir,
-            timeout=rt.runtime().get_option('general/0/git_timeout')
+            # FIXME: cast to float explicitly due to GH #2246
+            timeout=float(rt.runtime().get_option('general/0/git_timeout'))
         )
 
     @final