Merge branch 'develop' into anakin_test

CMakeLists.txt

@@ -65,6 +65,7 @@ option(REPLACE_ENFORCE_GLOG "Replace PADDLE_ENFORCE with glog/CHECK for better d
 option(WITH_ANAKIN      "Compile with Anakin library"                   OFF)
 option(WITH_GRPC     "Use grpc as the default rpc framework"            ${WITH_DISTRIBUTE})
 option(WITH_BRPC_RDMA     "Use brpc rdma as the rpc protocal"           OFF)
+option(WITH_INFERENCE    "Compile fluid inference library"              ON)
 option(WITH_SYSTEM_BLAS   "Use system blas library"           OFF)
 option(PY_VERSION       "Compile PaddlePaddle with python3 support"     ${PY_VERSION})
 

cmake/generic.cmake

@@ -264,6 +264,8 @@ function(cc_test TARGET_NAME)
              WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR})
     if (${cc_test_SERIAL})
         set_property(TEST ${TARGET_NAME} PROPERTY RUN_SERIAL 1)
+
+    set_property(TEST ${TARGET_NAME} PROPERTY ENVIRONMENT FLAGS_cpu_deterministic=true)
     set_property(TEST ${TARGET_NAME} PROPERTY ENVIRONMENT FLAGS_init_allocated_mem=true)
     set_property(TEST ${TARGET_NAME} PROPERTY ENVIRONMENT FLAGS_cudnn_deterministic=true)
     endif()
@@ -330,6 +332,8 @@ function(nv_test TARGET_NAME)
     add_test(${TARGET_NAME} ${TARGET_NAME})
     if (nv_test_SERIAL)
         set_property(TEST ${TARGET_NAME} PROPERTY RUN_SERIAL 1)
+
+    set_property(TEST ${TARGET_NAME} PROPERTY ENVIRONMENT FLAGS_cpu_deterministic=true)
     set_property(TEST ${TARGET_NAME} PROPERTY ENVIRONMENT FLAGS_init_allocated_mem=true)
     set_property(TEST ${TARGET_NAME} PROPERTY ENVIRONMENT FLAGS_cudnn_deterministic=true)
     endif()
@@ -580,6 +584,7 @@ function(py_test TARGET_NAME)
     cmake_parse_arguments(py_test "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
     add_test(NAME ${TARGET_NAME}
              COMMAND env FLAGS_init_allocated_mem=true FLAGS_cudnn_deterministic=true
+             FLAGS_cpu_deterministic=true
              PYTHONPATH=${PADDLE_BINARY_DIR}/python ${py_test_ENVS}
              ${PYTHON_EXECUTABLE} -u ${py_test_SRCS} ${py_test_ARGS}
              WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR})

paddle/fluid/CMakeLists.txt

@@ -5,5 +5,7 @@ add_subdirectory(operators)
 add_subdirectory(pybind)
 add_subdirectory(string)
 add_subdirectory(recordio)
-# NOTE: please add subdirectory inference at last.
-add_subdirectory(inference)
+if(WITH_INFERENCE)
+  # NOTE: please add subdirectory inference at last.
+  add_subdirectory(inference)
+endif()

paddle/fluid/framework/details/build_strategy.h

@@ -21,6 +21,26 @@ namespace framework {
 namespace details {
 
 struct BuildStrategy {
+  // ParallelExecutor supports two modes of ReduceStrategy, kAllReduce and
+  // kReduce, for CPU and GPU. If you use kAllReduce, different threads
+  // optimize their parameters separately. If you use kReduce, the optimizations
+  // of parameters are distributed to different threads.
+  // For example, a model has 100 parameters and is running with four threads,
+  // if you choose kAllReduce, every thread is to optimize 100 parameters
+  // separately, if you choose kReduce, every thread is to optimize 25
+  // parameters.
+  // Of particular note is, if you use kReduce when using CPU training,
+  // all the parameters are shared between different threads. This feature will
+  // save memory.
+  // FIXME(zcd): The result of the two modes(kAllReduce and kReduce) maybe not
+  // equal for GPU. Because, the result of the different order of summing maybe
+  // different, for example, the result of `a+b+c+d` may be different with the
+  // result of `c+a+b+d`.
+  // For GPU, the implementation of kAllReduce and kReduce is adopted NCCL,
+  // so the result of kAllReduce and kReduce maybe not equal.
+  // For CPU, if you want to fix the order of summing to make the result
+  // of kAllReduce and kReduce no diff, you can add
+  // `FLAGS_cpu_deterministic=true` to env.
   enum class ReduceStrategy { kAllReduce = 0, kReduce = 1 };
 
   enum class GradientScaleStrategy {

paddle/fluid/framework/details/reduce_op_handle.cc

@@ -18,6 +18,10 @@
 #include "paddle/fluid/framework/details/variable_visitor.h"
 #include "paddle/fluid/platform/profiler.h"
 
+DEFINE_bool(
+    cpu_deterministic, false,
+    "Whether to make the result of computation deterministic in CPU side.");
+
 namespace paddle {
 namespace framework {
 namespace details {
@@ -91,11 +95,33 @@ void ReduceOpHandle::RunImpl() {
   } else {
     std::vector<const LoDTensor *> lod_tensors =
         GetInputValues<LoDTensor>(in_var_handles, var_scopes);
+
     if (paddle::platform::is_cpu_place(lod_tensors[0]->place())) {
       this->RunAndRecordEvent([&] {
+        // FIXME(zcd): The order of summing is important,
+        // especially when the type of data is float or double.
+        // For example, the result of `a+b+c+d` may be different
+        // with the result of `c+a+b+d`, so the summing order should be fixed.
+        if (!FLAGS_cpu_deterministic) {
           ReduceLoDTensor func(lod_tensors,
                                out_var->GetMutable<framework::LoDTensor>());
           VisitDataType(ToDataType(lod_tensors[0]->type()), func);
+        } else {
+          // We sum lod_tensors to reduce_sum_trg which is in local_scopes_0
+          // here, but it doesn't mean reduce_sum_trg must be in local_scopes_0.
+          auto &reduce_sum_trg = *this->local_scopes_[0]
+                                      ->FindVar(kLocalExecScopeName)
+                                      ->Get<Scope *>()
+                                      ->FindVar(out_var_handle->name_)
+                                      ->GetMutable<framework::LoDTensor>();
+          ReduceLoDTensor func(lod_tensors, &reduce_sum_trg);
+          VisitDataType(ToDataType(lod_tensors[0]->type()), func);
+
+          auto trg = out_var->GetMutable<framework::LoDTensor>();
+          if (reduce_sum_trg.data<void>() != trg->data<void>()) {
+            TensorCopy(reduce_sum_trg, platform::CPUPlace(), trg);
+          }
+        }
       });
     } else if (paddle::platform::is_gpu_place(lod_tensors[0]->place())) {
 #ifdef PADDLE_WITH_CUDA

paddle/fluid/operators/read_op.cc

@@ -15,6 +15,7 @@
 #include "paddle/fluid/framework/op_registry.h"
 #include "paddle/fluid/framework/reader.h"
 #include "paddle/fluid/operators/detail/safe_ref.h"
+#include "paddle/fluid/platform/profiler.h"
 
 namespace paddle {
 namespace operators {
@@ -65,6 +66,12 @@ class ReadOp : public framework::OperatorBase {
             .GetMutable<framework::ReaderHolder>();
     std::vector<std::string> out_arg_names = Outputs("Out");
     std::vector<framework::LoDTensor> ins;
+
+    // For profiling
+    platform::DeviceContextPool& pool = platform::DeviceContextPool::Instance();
+    auto& ctx = *pool.Get(dev_place);
+    platform::RecordEvent record_event(Type(), &ctx);
+
     reader->ReadNext(&ins);
     if (ins.empty()) {
       if (Attr<bool>("throw_eof_exp")) {

python/paddle/fluid/__init__.py

@@ -123,7 +123,8 @@ def __bootstrap__():
     read_env_flags = [
         'use_pinned_memory', 'check_nan_inf', 'benchmark', 'warpctc_dir',
         'eager_delete_scope', 'use_mkldnn', 'initial_cpu_memory_in_mb',
-        'init_allocated_mem', 'free_idle_memory', 'paddle_num_threads'
+        'init_allocated_mem', 'free_idle_memory', 'paddle_num_threads',
+        'cpu_deterministic'
     ]
     if core.is_compiled_with_dist():
         read_env_flags.append('rpc_deadline')

python/paddle/fluid/tests/unittests/dist_se_resnext.py

@@ -174,6 +174,9 @@ class SE_ResNeXt():
             padding=(filter_size - 1) / 2,
             groups=groups,
             act=None,
+            # avoid pserver CPU init differs from GPU
+            param_attr=fluid.ParamAttr(
+                initializer=fluid.initializer.Constant()),
             bias_attr=False)
         return fluid.layers.batch_norm(input=conv, act=act)
 
@@ -194,10 +197,8 @@ class SE_ResNeXt():
 
 def get_model(batch_size):
     # Input data
-    image = fluid.layers.fill_constant(
-        shape=[batch_size, 3, 224, 224], dtype='float32', value=0.0)
-    label = fluid.layers.fill_constant(
-        shape=[batch_size, 1], dtype='int64', value=0.0)
+    image = fluid.layers.data(name="data", shape=[3, 224, 224], dtype='float32')
+    label = fluid.layers.data(name="int64", shape=[1], dtype='int64')
 
     # Train program
     model = SE_ResNeXt(layers=50)
@@ -222,8 +223,10 @@ def get_model(batch_size):
     lr = [base_lr * (0.1**i) for i in range(len(bd) + 1)]
 
     optimizer = fluid.optimizer.Momentum(
-        learning_rate=fluid.layers.piecewise_decay(
-            boundaries=bd, values=lr),
+        # FIXME(typhoonzero): add back LR decay once ParallelExecutor fixed.
+        #learning_rate=fluid.layers.piecewise_decay(
+        #    boundaries=bd, values=lr),
+        learning_rate=base_lr,
         momentum=0.9,
         regularization=fluid.regularizer.L2Decay(1e-4))
     optimizer.minimize(avg_cost)
@@ -232,7 +235,7 @@ def get_model(batch_size):
     train_reader = paddle.batch(
         paddle.dataset.flowers.train(), batch_size=batch_size)
     test_reader = paddle.batch(
-        paddle.dataset.flowers.test(), batch_size=batch_size)
+        paddle.dataset.flowers.test(use_xmap=False), batch_size=batch_size)
 
     return test_program, avg_cost, train_reader, test_reader, acc_top1, out
 
@@ -256,7 +259,6 @@ class DistSeResneXt2x2:
                            trainers)
         pserver_prog = t.get_pserver_program(current_endpoint)
         startup_prog = t.get_startup_program(current_endpoint, pserver_prog)
-
         place = fluid.CPUPlace()
         exe = fluid.Executor(place)
         exe.run(startup_prog)
@@ -302,12 +304,19 @@ class DistSeResneXt2x2:
         ]
 
         feeder = fluid.DataFeeder(feed_var_list, place)
-        reader_generator = train_reader()
-        first_loss, = exe.run(fetch_list=[avg_cost.name])
+        reader_generator = test_reader()
+
+        data = next(reader_generator)
+        first_loss, = exe.run(fetch_list=[avg_cost.name],
+                              feed=feeder.feed(data))
         print(first_loss)
+
         for i in xrange(5):
-            loss, = exe.run(fetch_list=[avg_cost.name])
-        last_loss, = exe.run(fetch_list=[avg_cost.name])
+            data = next(reader_generator)
+            loss, = exe.run(fetch_list=[avg_cost.name], feed=feeder.feed(data))
+
+        data = next(reader_generator)
+        last_loss, = exe.run(fetch_list=[avg_cost.name], feed=feeder.feed(data))
         print(last_loss)
 
 

python/paddle/fluid/tests/unittests/test_dist_base.py

@@ -63,7 +63,8 @@ class TestDistBase(unittest.TestCase):
             "PATH": os.getenv("PATH"),
             "PYTHONPATH": os.getenv("PYTHONPATH"),
             "LD_LIBRARY_PATH": os.getenv("LD_LIBRARY_PATH"),
-            "FLAGS_fraction_of_gpu_memory_to_use": "0.15"
+            "FLAGS_fraction_of_gpu_memory_to_use": "0.15",
+            "FLAGS_cudnn_deterministic": "1"
         }
         # Run local to get a base line
         env_local = {"CUDA_VISIBLE_DEVICES": "0"}

python/paddle/fluid/tests/unittests/test_dist_se_resnext.py

@@ -17,8 +17,7 @@ from test_dist_base import TestDistBase
 
 class TestDistSeResneXt2x2(TestDistBase):
     def test_se_resnext(self):
-        # TODO(paddle-dev): Is the delta too large?
-        self.check_with_place("dist_se_resnext.py", delta=0.2)
+        self.check_with_place("dist_se_resnext.py")
 
 
 if __name__ == "__main__":

python/paddle/fluid/tests/unittests/test_dist_transpiler.py

@@ -359,5 +359,110 @@ class TestL2DecayWithPiecewise(TranspilerTest):
             ["sum", "scale", "scale", "elementwise_add", "momentum"])
 
 
+class TestDistLookupTableBase(TranspilerTest):
+    def network_with_table(self, is_sparse, is_distributed):
+        def emb_pool(ids):
+            table_size = 1000
+            emb_size = 64
+            emb = fluid.layers.embedding(
+                input=ids,
+                size=[table_size, emb_size],
+                dtype='float32',
+                param_attr='shared_w',  # share parameter
+                is_sparse=is_sparse,
+                is_distributed=is_distributed)
+            pool = fluid.layers.sequence_pool(input=emb, pool_type='average')
+            return pool
+
+        title_ids = fluid.layers.data(
+            name='title_ids', shape=[1], dtype='int64', lod_level=1)
+        brand_ids = fluid.layers.data(
+            name='brand_ids', shape=[1], dtype='int64', lod_level=1)
+        title_emb = emb_pool(title_ids)
+        brand_emb = emb_pool(brand_ids)
+        fc0 = fluid.layers.concat(input=[title_emb, brand_emb], axis=1)
+        predict = fluid.layers.fc(input=fc0,
+                                  size=2,
+                                  act=None,
+                                  param_attr=fluid.ParamAttr(name='fc_w'),
+                                  bias_attr=fluid.ParamAttr(name='fc_b'))
+
+        label = fluid.layers.data(name='label', shape=[1], dtype='int64')
+        cost = fluid.layers.cross_entropy(input=predict, label=label)
+        avg_cost = fluid.layers.mean(cost)
+        optimizer = fluid.optimizer.Adam(learning_rate=0.003)
+        optimizer.minimize(avg_cost)
+
+
+class TestLocalLookupTable(TestDistLookupTableBase):
+    def net_conf(self):
+        self.network_with_table(is_sparse=True, is_distributed=False)
+
+    def transpiler_test_impl(self):
+        pserver1, startup1 = self.get_pserver(self.pserver1_ep)
+
+        self.assertEqual(len(pserver1.blocks), 3)
+        # 0 listen_and_serv
+        # 1 optimize for fc_w or fc_b adam
+        self.assertEqual([op.type for op in pserver1.blocks[1].ops],
+                         ["sum", "scale", "adam", "scale", "scale"])
+        # 2 optimize for table adam
+        # NOTE: if param is not selected rows, the grad will scaled to grad / trainer_num
+        self.assertEqual([op.type for op in pserver1.blocks[2].ops],
+                         ["sum", "adam", "scale", "scale"])
+
+        trainer = self.get_trainer()
+        self.assertEqual(len(trainer.blocks), 1)
+        ops = [
+            'lookup_table', 'sequence_pool', 'lookup_table', 'sequence_pool',
+            'concat', 'mul', 'elementwise_add', 'cross_entropy', 'mean',
+            'fill_constant', 'mean_grad', 'cross_entropy_grad',
+            'elementwise_add_grad', 'send', 'mul_grad', 'send', 'concat_grad',
+            'sequence_pool_grad', 'lookup_table_grad', 'sequence_pool_grad',
+            'lookup_table_grad', 'sum', 'split_selected_rows', 'send',
+            'send_barrier', 'recv', 'recv', 'recv', 'fetch_barrier', 'concat'
+        ]
+        self.assertEqual([op.type for op in trainer.blocks[0].ops], ops)
+
+
+class TestDistLookupTable(TestDistLookupTableBase):
+    def net_conf(self):
+        self.network_with_table(is_sparse=True, is_distributed=True)
+
+    def transpiler_test_impl(self):
+        pserver1, startup1 = self.get_pserver(self.pserver1_ep)
+
+        self.assertEqual(len(pserver1.blocks), 6)
+        # 0 listen_and_serv
+        # 1 optimize for fc_w or fc_b adam
+        self.assertEqual([op.type for op in pserver1.blocks[1].ops],
+                         ["sum", "scale", "adam", "scale", "scale"])
+        # 2 optimize for table sgd
+        self.assertEqual([op.type for op in pserver1.blocks[2].ops],
+                         ["sum", "sgd"])
+        # 3 prefetch -> lookup_sparse_table for data0
+        self.assertEqual([op.type for op in pserver1.blocks[3].ops],
+                         ["lookup_sparse_table"])
+        # 4 prefetch -> lookup_sparse_table for data1
+        self.assertEqual([op.type for op in pserver1.blocks[4].ops],
+                         ["lookup_sparse_table"])
+        # 5 save table
+        self.assertEqual([op.type for op in pserver1.blocks[5].ops], ["save"])
+
+        trainer = self.get_trainer()
+        self.assertEqual(len(trainer.blocks), 1)
+        ops = [
+            'split_ids', 'prefetch', 'merge_ids', 'sequence_pool', 'split_ids',
+            'prefetch', 'merge_ids', 'sequence_pool', 'concat', 'mul',
+            'elementwise_add', 'cross_entropy', 'mean', 'fill_constant',
+            'mean_grad', 'cross_entropy_grad', 'elementwise_add_grad', 'send',
+            'mul_grad', 'send', 'concat_grad', 'sequence_pool_grad',
+            'lookup_table_grad', 'sequence_pool_grad', 'lookup_table_grad',
+            'sum', 'split_ids', 'send', 'send_barrier', 'recv', 'recv',
+            'fetch_barrier'
+        ]
+        self.assertEqual([op.type for op in trainer.blocks[0].ops], ops)
+
+
 if __name__ == "__main__":
     unittest.main()

python/paddle/fluid/tests/unittests/test_parallel_executor_seresnext.py

@@ -198,7 +198,7 @@ class TestResnet(TestParallelExecutorBase):
                                       model,
                                       use_cuda,
                                       iter=20,
-                                      delta2=1e-4):
+                                      delta2=1e-6):
         if use_cuda and not core.is_compiled_with_cuda():
             return
 
@@ -276,10 +276,10 @@ class TestResnet(TestParallelExecutorBase):
             model=SE_ResNeXt50Small, use_cuda=False, iter=2, delta2=1e-3)
 
     def test_seresnext_with_new_strategy(self):
-        # self._compare_reduce_and_allreduce(
-        #     model=SE_ResNeXt50Small, use_cuda=True)
         self._compare_reduce_and_allreduce(
-            model=SE_ResNeXt50Small, use_cuda=False, iter=5, delta2=1e-2)
+            model=SE_ResNeXt50Small, use_cuda=True, delta2=1e-2)
+        self._compare_reduce_and_allreduce(
+            model=SE_ResNeXt50Small, use_cuda=False, iter=5)
 
 
 if __name__ == '__main__':

python/paddle/fluid/transpiler/distribute_transpiler.py

@@ -896,8 +896,6 @@ class DistributeTranspiler(object):
             self.table_name
         ][0]
         table_opt_block = pserver_program.create_block(pre_block_idx)
-        # only support sgd now
-        assert table_opt_op.type == "sgd"
 
         if self.sync_mode:
             # create grad vars in pserver program
@@ -937,11 +935,12 @@ class DistributeTranspiler(object):
             "LearningRate": [lr_var]
         }
         outputs = {"ParamOut": [param_var]}
-        table_opt_block.append_op(
-            type=table_opt_op.type,
-            inputs=inputs,
-            outputs=outputs,
-            attrs=table_opt_op.attrs)
+        # only support sgd now
+        import logging
+        logging.warn(
+            "distribute lookup table only support sgd optimizer, change it's optimizer to sgd instead of "
+            + table_opt_op.type)
+        table_opt_block.append_op(type="sgd", inputs=inputs, outputs=outputs)
 
         # add table parameter gradient and it's block id to grad_to_block_id
         grad_to_block_id.append(grad_var.name + ":" + str(table_opt_block.idx))