[MXNET-#16795] Byteps-KVStore: Intergrate Byteps into mxnet as new ty…

…pe of kvstore backend (#17555) * Add Byteps backend for kvstore * Add a temp launcher for byteps backend * make the test fit for byteps kvstore. * final workable test * Remove trashy print and logs * correct comment * add hostfile for ci test * add ci test for byteps kvstore * add visibile devices for byteps-kvstore ci test * add licenses for tools/byteps_launcher.py * syntax error * pylint error (remove unused import like logging) * pylint error * pylint error * enable launching without hostfile (local byteps) * 1. rename byteps_kvstore.py to byteps.py; 2. shorten the launch option to ; 3. add instruction for -H and -SH options for launch; 4. add documentation for byteps kvstore in kvstore/base.py: create(name='local') * edit documentation of KVStoreBase::is_capable(capability); reture fasle for BytePS(KVStoreBase):is_capable(any). * pylint error * remove an error of arg.byteps * use --env option to set workers' environment * error in byteps-launcher.py * remove the unpurposed editing mistake in runtime_functions.sh * disable cpu support for byteps kvstore. * 1. format the document to avoid julia doc build error; 2. little change to nightly test; 3. add byteps copy right declararation in byteps_launcher.py 4. if args.byteps == True ===> if args.byteps * remove the --scheduler_ip and --scheduler_port options in launch.py * 1. maintain the origin value of broadcast and pushpull 2. optimize when out = value or [out]=value 3. add some missing documentation to avoid doc building error. * Add bytePS to CI * add dependency * +integrationtest_ubuntu_gpu_byteps * add byteps pipeline * disable a few tests * remove more tests * fix permission * remove apt-get * fix python path * improve logging * fix printns * add back CI Co-authored-by: Ubuntu <[email protected]> Co-authored-by: Piyush Ghai <[email protected]> Co-authored-by: eric-haibin-lin <[email protected]> Co-authored-by: eric-haibin-lin <--global> Co-authored-by: Lin <[email protected]>
apache · Apr 7, 2020 · c244f9f · c244f9f
1 parent ff234db
commit c244f9f
Show file tree

Hide file tree

Showing 10 changed files with 624 additions and 3 deletions.
diff --git a/ci/docker/runtime_functions.sh b/ci/docker/runtime_functions.sh
@@ -1004,6 +1004,7 @@ cd_unittest_ubuntu() {
         # Adding these here as CI doesn't test all CUDA environments
         $python_cmd example/image-classification/test_score.py
         integrationtest_ubuntu_gpu_dist_kvstore
+        integrationtest_ubuntu_gpu_byteps
     fi
 
     if [[ ${mxnet_variant} = *mkl ]]; then
@@ -1351,6 +1352,24 @@ integrationtest_ubuntu_gpu_dist_kvstore() {
     popd
 }
 
+integrationtest_ubuntu_gpu_byteps() {
+    set -ex
+    pushd .
+    export PYTHONPATH=$PWD/python/
+    export BYTEPS_WITHOUT_PYTORCH=1
+    export BYTEPS_WITHOUT_TENSORFLOW=1
+    git clone -b v0.2 https://github.com/bytedance/byteps/ --recursive
+    cd byteps && python3 setup.py install --user && cd -
+
+    export MXNET_STORAGE_FALLBACK_LOG_VERBOSE=0
+    export MXNET_SUBGRAPH_VERBOSE=0
+    export DMLC_LOG_STACK_TRACE_DEPTH=10
+    cd tests/nightly/
+    python3 ../../tools/launch.py -n 1 -s 1 --byteps --env NVIDIA_VISIBLE_DEVICES:0,1 python3 dist_device_sync_kvstore_byteps.py
+    popd
+}
+
+
 test_ubuntu_cpu_python3() {
     set -ex
     pushd .

diff --git a/ci/jenkins/Jenkins_steps.groovy b/ci/jenkins/Jenkins_steps.groovy
@@ -1277,6 +1277,20 @@ def test_unix_distributed_kvstore_cpu() {
     }]
 }
 
+def test_unix_byteps_gpu() {
+    return ['byteps tests GPU': {
+      node(NODE_LINUX_GPU) {
+        ws('workspace/it-byteps') {
+          timeout(time: max_time, unit: 'MINUTES') {
+            utils.unpack_and_init('gpu', mx_lib)
+            utils.docker_run('ubuntu_gpu_cu101', 'integrationtest_ubuntu_gpu_byteps', true)
+            utils.publish_test_coverage()
+          }
+        }
+      }
+    }]
+}
+
 def test_unix_distributed_kvstore_gpu() {
     return ['dist-kvstore tests GPU': {
       node(NODE_LINUX_GPU) {

diff --git a/ci/jenkins/Jenkinsfile_edge b/ci/jenkins/Jenkinsfile_edge
@@ -44,7 +44,7 @@ core_logic: {
 
   utils.parallel_stage('Tests', [
     custom_steps.test_qemu_armv7_cpu()
-  ]) 
+  ])
 }
 ,
 failure_handler: {

diff --git a/ci/jenkins/Jenkinsfile_unix_gpu b/ci/jenkins/Jenkinsfile_unix_gpu
@@ -60,6 +60,7 @@ core_logic: {
     custom_steps.test_unix_cpp_package_gpu(),
     custom_steps.test_unix_scala_gpu(),
     custom_steps.test_unix_distributed_kvstore_gpu(),
+    custom_steps.test_unix_byteps_gpu(),
     custom_steps.test_static_python_gpu(),
     custom_steps.test_static_python_gpu_cmake(),
     custom_steps.test_unix_python3_gpu_no_tvm_op(),

diff --git a/python/mxnet/kvstore/__init__.py b/python/mxnet/kvstore/__init__.py
@@ -22,3 +22,4 @@
 from .kvstore import *
 from .base import *
 from .kvstore_server import *
+from .byteps import *
diff --git a/python/mxnet/kvstore/base.py b/python/mxnet/kvstore/base.py
@@ -313,6 +313,8 @@ def pushpull(self, key, value, out=None, priority=0):
     def is_capable(capability):
         """Queries if the KVStore type supports certain capability, such as optimizer algorithm,
         gradient compression, sparsity, etc.
+        If the kvstore does not store weights in server part, then no optimizer is supported,
+        this function will return False.
 
         Parameters
         ----------
@@ -427,10 +429,15 @@ def create(name='local'):
     No two updates happen on the same weight at the same time. However, the order is not
     guaranteed.
 
+    ``byteps``: Use byteps as broadcast/pushpull backend.
+    This kind of kvstore doesn't store weights, thus there won't be optimizer in this kvstore server.
+    Byteps doesn't support pure cpu training, so be sure to enable gpu training when using this kvstore.
+
     Parameters
     ----------
-    name : {'local', 'device', 'nccl', 'dist_sync', 'dist_device_sync', 'dist_async', 'horovod'}
+    name : {'local', 'device', 'nccl', 'dist_sync', 'dist_device_sync', 'dist_async', 'horovod', 'byteps'}
         The type of KVStore.
+
     Returns
     -------
     kv : KVStoreBase

diff --git a/python/mxnet/kvstore/byteps.py b/python/mxnet/kvstore/byteps.py
@@ -0,0 +1,255 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements.  See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership.  The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License.  You may obtain a copy of the License at
+#
+#   http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied.  See the License for the
+# specific language governing permissions and limitations
+# under the License.
+
+# coding: utf-8
+""" BytePS backend for MXNet KVStore"""
+from __future__ import absolute_import
+
+from ..ndarray import NDArray
+from .base import KVStoreBase
+
+__all__ = ['BytePS']
+
+
+@KVStoreBase.register
+class BytePS(KVStoreBase):
+    """BytePS backend for MXNet KVStore interface."""
+
+    def __init__(self):
+        """Initializes a new KVStore."""
+        try:
+            import byteps.mxnet as bps
+            self.handle = bps
+        except ImportError as err:
+            print('Did not find BytePS library. Please install BytePS first')
+            raise err
+        self.handle.init()
+
+    def broadcast(self, key, value, out, priority=0):
+        """ Broadcast the value NDArray at rank 0 to all ranks' out. If out is None,
+        the result is stored in `value`.
+
+        Parameters
+        ----------
+        key : str, or int
+            The keys.
+        value : NDArray, or list of NDArray
+            Values corresponding to the key.
+        out : NDArray, or lise of NDArray
+            Values corresponding to the keys.
+
+        Examples
+        --------
+        >>> # broadcast a single key-value pair
+        >>> shape = (2,3)
+        >>> kv = mx.kv.create('byteps')
+        >>> a = mx.nd.zeros(shape)
+        >>> kv.broadcast('3', mx.nd.ones(shape)*2, out=a)
+        >>> print a.asnumpy()
+        [[ 2.  2.  2.]
+        [ 2.  2.  2.]]
+        """
+        # do not accept list or tuple for key/value
+        assert isinstance(key, (str, int))
+
+        # unpack the list if it contains just one NDArray
+        value = value[0] if isinstance(
+            value, list) and len(value) == 1 else value
+        assert isinstance(
+            value, NDArray), "The type of value can only be NDArray or list of NDArray which has only one element."
+        assert value.context.device_type == 'gpu', "Byteps KVStore only support GPU context for broadcast value."
+
+        # optimzation when out = value or out = [value]
+        if isinstance(out, (list, tuple)) and len(out) == 1:
+            inplace = value is out[0]
+        else:
+            inplace = value is out
+
+        if inplace:
+            broadcast_value = value
+        else:
+            broadcast_value = value.copy()
+        # for non-root-rank, assign value with 0, thus the result of pushpull will be
+        # equal to the value of root-rank, thus implementing broadcast.
+        root_rank = 0
+        if self.rank != root_rank:
+            broadcast_value.__imul__(0)
+        self.handle.byteps_push_pull(broadcast_value, version=0, priority=priority,
+                                     name=str(key), is_average=False)
+        # Make sure tensors pushed to MXNet engine get processed such that all
+        # workers are synced before starting training.
+        broadcast_value.wait_to_read()
+
+        out = out if isinstance(out, list) else [out]
+        for o in out:
+            broadcast_value.copyto(o)
+
+    def pushpull(self, key, value, out=None, priority=0):
+        """ Performs push and pull a single value from the store.
+        This function is coalesced form of push and pull operations.
+        `value` is pushed to the kvstore server for the specified keys and the aggregated
+        values are pulled from the server to `out`. If `out` is not specified the pulled
+        values are written to `value`.
+
+        Parameters
+        ----------
+        key : str, or int
+            The key.
+        value : NDArray, or list of NDArray
+            Values corresponding to the key.
+        out: NDArray, or list of NDArray
+            Values corresponding to the key.
+        priority : int, optional
+            The priority of the operation.
+            Higher priority operations are likely to be executed before other actions.
+
+        Examples
+        --------
+        >>> # pushpull a single key-value pair
+        >>> kv.pushpull('3', mx.nd.ones(shape)*8, out=a)
+        >>> print a.asnumpy()
+        [[ 8.  8.  8.]
+        [ 8.  8.  8.]]
+        """
+        # the most common operation operates on one NDArray as `value`, and
+        # `out` is set to None, for inplace pushpull.
+
+        assert isinstance(key, (str, int))
+
+        # unpack the list if it contains just one NDArray
+        value = value[0] if isinstance(
+            value, list) and len(value) == 1 else value
+        assert isinstance(
+            value, NDArray), "The type of value can only be NDArray or list of NDArray which has only one element."
+        assert value.context.device_type == 'gpu', "Byteps KVStore only support GPU context for pushpull value"
+
+        # optimzation when out = value or out = [value]
+        if isinstance(out, (list, tuple)) and len(out) == 1:
+            inplace = value is out[0]
+        else:
+            inplace = value is out
+
+        if inplace:
+            pushpull_value = value
+        else:
+            pushpull_value = value.copy()
+
+        self.handle.byteps_push_pull(pushpull_value, version=0, priority=priority,
+                                     name=str(key), is_average=False)
+
+        if out is not None:
+            out = out if isinstance(out, list) else [out]
+            for o in out:
+                pushpull_value.copyto(o)
+
+    @staticmethod
+    def is_capable(capability):
+        """Queries if the KVStore type supports certain capability, such as optimizer algorithm,
+        gradient compression, sparsity, etc.
+        As byteps server does not store weight, this function will return false for any capabilities.
+
+        Parameters
+        ----------
+        capability: str
+            The capability to query
+
+        Returns
+        -------
+        result : bool
+            Whether the capability is supported or not.
+        """
+        return False
+
+    @property
+    def type(self):
+        """ Returns the type of this kvstore.
+
+        Returns
+        -------
+        type : str
+            the string type
+        """
+        return 'byteps'
+
+    @property
+    def local_rank(self):
+        """ Returns the local rank of this worker on the node.
+
+        Returns
+        -------
+        rank : int
+            The local rank of this node, which is in range [0, num_workers_on_current_node())
+        """
+        return self.handle.local_rank()
+
+    @property
+    def rank(self):
+        """ Returns the rank of this worker node.
+
+        Returns
+        -------
+        rank : int
+            The rank of this node, which is in range [0, num_workers())
+        """
+        return self.handle.rank()
+
+    @property
+    def num_workers(self):
+        """Returns the number of worker nodes.
+
+        Returns
+        -------
+        size :int
+            The number of worker nodes.
+        """
+        return self.handle.size()
+
+    def set_optimizer(self, optimizer):
+        """
+        Not Implement yet.
+
+        Parameters
+        ----------
+        optimizer : KVStoreBase
+            The new optimizer for the store
+        """
+        raise NotImplementedError()
+
+    def save_optimizer_states(self, fname, dump_optimizer=False):
+        """
+        Not Implement yet.
+
+        Parameters
+        ----------
+        fname : str
+            Path to the output states file.
+        dump_optimizer : bool, default False
+            Whether to also save the optimizer itself. This would also save optimizer
+            information such as learning rate and weight decay schedules.
+        """
+        raise NotImplementedError()
+
+    def load_optimizer_states(self, fname):
+        """
+        Not Implement yet.
+
+        Parameters
+        ----------
+        fname : str
+            Path to input states file.
+        """
+        raise NotImplementedError()