Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into add-merge-splited-ids

cmake/configure.cmake

@@ -118,6 +118,10 @@ endif()
 set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${SIMD_FLAG}")
 set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${SIMD_FLAG}")
 
+if(WITH_DISTRIBUTE)
+  add_definitions(-DPADDLE_WITH_DISTRIBUTE)
+endif()
+
 if(WITH_GOLANG)
   # we need to symlink Paddle directory into GOPATH. If we
   # don't do it and we have code that depends on Paddle, go

doc/survey/dynamic_graph.md

+# Automatic Differentiation with the Tape
+
+## Automatic Differentiation
+
+A key challenge in the field of deep learning is to automatically derive the backward pass from the forward pass described algorithmically by researchers.  Such a derivation, or a transformation of the forward pass program, has been long studied before the recent prosperity of deep learning in the field known as [automatic differentiation](https://arxiv.org/pdf/1502.05767.pdf).
+
+## The Tape
+
+Given the forward pass program (usually in Python in practices), there are two strategies to derive the backward pass:
+
+1. from the forward pass program itself, or
+1. from the execution trace of the forward pass program, which is often known as the *tape*.
+
+This article surveys systems that follow the latter strategy.
+
+## Dynamic Network
+
+When we train a deep learning model, the tape changes every iteration as the input data change, so we have to re-derive the backward pass every iteration.  This is known as *dynamic network*.
+
+Deep learning systems that utilize the idea of dynamic network gained their popularities in recent years.  This article surveys two representative systems: [PyTorch](https://pytorch.org/) and [DyNet](https://dynet.readthedocs.io/en/latest/).
+
+## An Overview
+
+Both frameworks record a ‘tape’ of the computation and interpreting (or run-time compiling) a transformation of the tape played back in reverse. This tape is a different kind of entity than the original program.[[link]](http://www.bcl.hamilton.ie/~barak/papers/toplas-reverse.pdf)
+
+Consider the following code feedforward model.
+
+```python
+x = Variable(randn(20, 1)))
+label = Variable(randint(1))
+W_1, W_2 = Variable(randn(20, 20)), Variable(randn(10, 20))
+h = matmul(W_1, x)
+pred = matmul(W_2, x)
+loss = softmax(pred, label)
+loss.backward()
+```
+
+### 1) Dynet uses List to encode the Tape
+
+During the forward execution, a list of operators, in this case `matmul`, `matmul` and `softmax`, are recorded in the tape, along with the necessary information needed to do the backward such as pointers to the inputs and outputs. Then the tape is played in reverse order at `loss.backward()`.
+
+<details> 
+<summary></summary>
+digraph g {
+    graph [
+        rankdir = "LR"
+    ];
+    node [
+        fontsize = "16"
+        shape = "ellipse"
+    ];
+    edge [];
+    "node0" [
+        label = "<f0> type: matmul | <f1> input: W_1, x | <f2> output: h"
+        shape = "record"
+    ];
+    "node1" [
+        label = "<f0> type: matmul | <f1> input: W_2, h | <f2> output: pred"
+        shape = "record"
+    ];
+    "node2" [
+        label = "<f0> type: softmax | <f1> input: pred, label | <f2> output: loss"
+        shape = "record"
+    ];
+    "node0":f0 -> "node1":f0 [];
+    "node1":f0 -> "node2":f0 [];
+}
+</details>
+
+![Alt text](https://g.gravizo.com/svg?digraph%20g%20{%20graph%20[%20rankdir%20=%20%22LR%22%20];%20node%20[%20fontsize%20=%20%2216%22%20shape%20=%20%22ellipse%22%20];%20edge%20[];%20%22node0%22%20[%20label%20=%20%22%3Cf0%3E%20type:%20matmul%20|%20%3Cf1%3E%20input:%20W_1,%20x%20|%20%3Cf2%3E%20output:%20h%22%20shape%20=%20%22record%22%20];%20%22node1%22%20[%20label%20=%20%22%3Cf0%3E%20type:%20matmul%20|%20%3Cf1%3E%20input:%20W_2,%20h%20|%20%3Cf2%3E%20output:%20pred%22%20shape%20=%20%22record%22%20];%20%22node2%22%20[%20label%20=%20%22%3Cf0%3E%20type:%20softmax%20|%20%3Cf1%3E%20input:%20pred,%20label%20|%20%3Cf2%3E%20output:%20loss%22%20shape%20=%20%22record%22%20];%20%22node0%22:f0%20-%3E%20%22node1%22:f0%20[%20id%20=%200%20];%20%22node1%22:f0%20-%3E%20%22node2%22:f0%20[%20id%20=%201%20];%20})
+
+### 2) Pytorch uses Node Graph to encode the Tape
+
+The graph is composed of `Variable`s and `Function`s. During the forward execution, a `Variable` records its creator function, e.g. `h.creator = matmul`. And a Function records its inputs' previous/dependent functions `prev_func` through `creator`, e.g. `matmul.prev_func = matmul1`. At `loss.backward()`, a topological sort is performed on all `prev_func`s. Then the grad op is performed by the sorted order.
+
+<details> 
+<summary></summary>
+digraph g {
+    graph [
+        rankdir = "LR"
+    ];
+    
+    subgraph function {
+        node [
+            fontsize = "16"
+            style = filled
+            shape = "record"
+        ];
+        "matmul0" [ label = "<f0> type: matmul | prev_func: None" ];
+        "matmul1" [ label = "<f0> type: matmul | prev_func: matmul" ];
+        "softmax" [ label = "<f0> type: softmax | prev_func: matmul" ];
+    }
+    
+    subgraph variable {
+        node [
+            fontsize = "16"
+            shape = "Mrecord"
+            style = filled
+            fillcolor = white
+        ];
+        "x" [ label = "<f0> x | <f1> creator: None" ];
+        "label" [ label = "<f0> label | <f1> creator: None" ];
+        "W_1" [ label = "<f0> W_1 | <f1> creator: None" ];
+        "W_2" [ label = "<f0> W_2 | <f1> creator: None" ];
+        "h" [ label = "<f0> h | <f1> creator: None" ];
+        "pred" [ label = "<f0> pred | <f1> creator: matmul" ];
+        "loss" [ label = "<f0> loss | <f1> creator: softmax" ];
+    }
+    
+    subgraph data_flow {
+        "x":f0 -> "matmul0":f0;
+        "W_1":f0 -> "matmul0":f0;
+        "matmul0":f0 -> "h":f0;
+    
+        "h":f0 -> "matmul1":f0;
+        "W_2":f0 -> "matmul1":f0;
+        "matmul1":f0 -> "pred":f0;
+    
+        "pred":f0 -> "softmax":f0;
+        "label":f0 -> "softmax":f0;
+        "softmax":f0 -> "loss":f0;
+    }
+
+    subgraph prev_func {
+        edge [color="red", arrowsize="0.6", penwidth="1", constraint=false];
+        "matmul1":f1 -> "matmul0":f0;
+        "softmax":f1 -> "matmul1":f0;
+        label = "prev_func";
+    }
+}
+</details>
+
+![Alt text](https://g.gravizo.com/svg?digraph%20g%20{%20graph%20[%20rankdir%20=%20%22LR%22%20];%20subgraph%20function%20{%20node%20[%20fontsize%20=%20%2216%22%20style%20=%20filled%20shape%20=%20%22record%22%20];%20%22matmul0%22%20[%20label%20=%20%22%3Cf0%3E%20type:%20matmul%20|%20prev_func:%20None%22%20];%20%22matmul1%22%20[%20label%20=%20%22%3Cf0%3E%20type:%20matmul%20|%20prev_func:%20matmul%22%20];%20%22softmax%22%20[%20label%20=%20%22%3Cf0%3E%20type:%20softmax%20|%20prev_func:%20matmul%22%20];%20}%20subgraph%20variable%20{%20node%20[%20fontsize%20=%20%2216%22%20shape%20=%20%22Mrecord%22%20style%20=%20filled%20fillcolor%20=%20white%20];%20%22x%22%20[%20label%20=%20%22%3Cf0%3E%20x%20|%20%3Cf1%3E%20creator:%20None%22%20];%20%22label%22%20[%20label%20=%20%22%3Cf0%3E%20label%20|%20%3Cf1%3E%20creator:%20None%22%20];%20%22W_1%22%20[%20label%20=%20%22%3Cf0%3E%20W_1%20|%20%3Cf1%3E%20creator:%20None%22%20];%20%22W_2%22%20[%20label%20=%20%22%3Cf0%3E%20W_2%20|%20%3Cf1%3E%20creator:%20None%22%20];%20%22h%22%20[%20label%20=%20%22%3Cf0%3E%20h%20|%20%3Cf1%3E%20creator:%20None%22%20];%20%22pred%22%20[%20label%20=%20%22%3Cf0%3E%20pred%20|%20%3Cf1%3E%20creator:%20matmul%22%20];%20%22loss%22%20[%20label%20=%20%22%3Cf0%3E%20loss%20|%20%3Cf1%3E%20creator:%20softmax%22%20];%20}%20subgraph%20data_flow%20{%20%22x%22:f0%20-%3E%20%22matmul0%22:f0;%20%22W_1%22:f0%20-%3E%20%22matmul0%22:f0;%20%22matmul0%22:f0%20-%3E%20%22h%22:f0;%20%22h%22:f0%20-%3E%20%22matmul1%22:f0;%20%22W_2%22:f0%20-%3E%20%22matmul1%22:f0;%20%22matmul1%22:f0%20-%3E%20%22pred%22:f0;%20%22pred%22:f0%20-%3E%20%22softmax%22:f0;%20%22label%22:f0%20-%3E%20%22softmax%22:f0;%20%22softmax%22:f0%20-%3E%20%22loss%22:f0;%20}%20subgraph%20prev_func%20{%20edge%20[color=%22red%22,%20arrowsize=%220.6%22,%20penwidth=%221%22,%20constraint=false];%20%22matmul1%22:f1%20-%3E%20%22matmul0%22:f0;%20%22softmax%22:f1%20-%3E%20%22matmul1%22:f0;%20label%20=%20%22prev_func%22;%20}%20})
+
+Chainer and Autograd uses the similar techniques to record the forward pass. For details please refer to the appendix.
+
+## Design choices
+
+### 1) Dynet's List vs Pytorch's Node Graph
+
+What's good about List:
+1. It avoids a topological sort. One only needs to traverse the list of operators in reverse and calling the corresponding backward operator.
+1. It promises effient data parallelism implementations. One could count the time of usage of a certain variable during the construction list. Then in the play back, one knows the calculation of a variable has completed. This enables communication and computation overlapping.
+
+What's good about Node Graph:
+1. More flexibility. PyTorch users can mix and match independent graphs however they like, in whatever threads they like (without explicit synchronization). An added benefit of structuring graphs this way is that when a portion of the graph becomes dead, it is automatically freed. [[2]](https://openreview.net/pdf?id=BJJsrmfCZ) Consider the following example, Pytorch only does backward on SmallNet while Dynet does both BigNet and SmallNet.
+```python
+result = BigNet(data)
+loss = SmallNet(data)
+loss.backward()
+```
+
+### 2) Dynet's Lazy evaluation vs Pytorch's Immediate evaluation
+
+Dynet builds the list in a symbolic matter. Consider the following example
+```python
+for epoch in range(num_epochs):
+    for in_words, out_label in training_data:
+        dy.renew_cg()
+        W = dy.parameter(W_p)
+        b = dy.parameter(b_p)
+        score_sym = dy.softmax(W*dy.concatenate([E[in_words[0]],E[in_words[1]]])+b)
+        loss_sym = dy.pickneglogsoftmax(score_sym, out_label)
+        loss_val = loss_sym.value()
+        loss_sym.backward()
+```
+The computation of `lookup`, `concat`, `matmul` and `softmax` didn't happen until the call of `loss_sym.value()`. This defered execution is useful because it allows some graph-like optimization possible, e.g. kernel fusion.
+
+Pytorch chooses immediate evaluation. It avoids ever materializing a "forward graph"/"tape" (no need to explicitly call `dy.renew_cg()` to reset the list), recording only what is necessary to differentiate the computation, i.e. `creator` and `prev_func`.
+
+
+## What can fluid learn from them?
+
+TBD
+
+# Appendix
+
+### Overview
+
+| Framework | Has Tape | Core in C++ | First Release Date |
+|-----------|----------|-------------|--------------------|
+| Autograd  | No       | No          | Mar 5, 2015        |
+| Chainer   | No       | No          | Jun 5, 2015        |
+| Pytorch   | No       | Yes         | Aug 31, 2016       |
+| Dynet     | Yes      | Yes         | Oct 12, 2016       |
+
+### Source Code
+#### Autograd
+[Backward code](https://github.com/HIPS/autograd/blob/442205dfefe407beffb33550846434baa90c4de7/autograd/core.py#L8-L40). In the forward pass, a graph of VJPNode is constructed.
+```python
+# User API
+def make_grad(fun, x):
+    start_node = VJPNode.new_root()
+    end_value, end_node =  trace(start_node, fun, x)
+    return backward_pass(g, end_node), end_value
+
+# trace the forward pass by creating VJPNodes
+def trace(start_node, fun, x):
+    with trace_stack.new_trace() as t:
+        start_box = new_box(x, t, start_node)
+        end_box = fun(start_box)
+        return end_box._value, end_box._node
+
+def backward_pass(g, end_node):
+    outgrads = {end_node : (g, False)}
+    for node in toposort(end_node):
+        outgrad = outgrads.pop(node)
+        ingrads = node.vjp(outgrad[0])
+        for parent, ingrad in zip(node.parents, ingrads):
+            outgrads[parent] = add_outgrads(outgrads.get(parent), ingrad)
+    return outgrad[0]
+
+# Every VJPNode corresponds to a op_grad
+class VJPNode(Node):
+    __slots__ = ['parents', 'vjp']
+    def __init__(self, value, fun, args, kwargs, parent_argnums, parents):
+        self.parents = parents
+        vjpmaker = primitive_vjps[fun]
+        self.vjp = vjpmaker(parent_argnums, value, args, kwargs)
+```
+#### Chainer
+Example Code
+```python
+# (1) Function Set definition, creates FunctionNode
+model = FunctionSet(
+    l1=F.Linear(784, 100),
+    l2=F.Linear(100, 100),
+    l3=F.Linear(100, 10)).to_gpu()
+
+# (2) Optimizer Setup
+opt = optimizers.SGD()
+opt.setup(model)
+
+# (3) Forward computation
+def forward(x, t):
+    h1 = F.relu(model.l1(x))
+    h2 = F.relu(model.l2(h1))
+    y = model.l3(h2)
+    return F.softmax_cross_entropy(y, t)
+
+# (4) Training loop
+for epoch in xrange(n_epoch):
+    for i in xrange(0, N, b_size):
+        x = Variable(to_gpu(...))
+        t = Variable(to_gpu(...))
+        opt.zero_grads()
+        loss = forward(x, t)
+        loss.backward()
+        opt.update()
+```
+In `forward(x, t)`, a graph of [`VariableNode`](https://github.com/chainer/chainer/blob/master/chainer/variable.py#L110) and [`FunctionNode`](https://github.com/chainer/chainer/blob/a69103a4aa59d5b318f39b01dbcb858d465b89cf/chainer/function_node.py#L19) is constructed. Every output's `VariableNode.creator` is pointed to the `FunctionNode`.
+```python
+class FunctionNode(object):
+    ...
+    def apply(self, inputs):
+        outputs = self.forward(inputs)
+        ret = tuple([variable.Variable(y, requires_grad=requires_grad)
+                     for y in outputs])
+        # Topological ordering
+        self.rank = max([x.rank for x in inputs]) if input_vars else 0
+        # Add backward edges
+        for y in ret:
+            y.creator_node = self
+        self.inputs = tuple([x.node for x in input_vars])
+        self.outputs = tuple([y.node for y in ret])
+
+        return ret
+```
+`loss.backward()` will calculate the accumulated gradient of all variables. All the backward of `FunctionNode`s will be called based on the topological order.
+```python
+class VariableNode(object):
+    ...
+    def backward(self, retain_grad, loss_scale):
+        if self.creator_node is None:
+            return
+
+        cand_funcs = []
+        seen_set = set()
+        grads = {}
+
+        # Initialize error by 1, if this is a loss variable
+        if self.data.size == 1 and self._grad_var is None:
+            self.grad = numpy.ones_like(self.data)
+        grads[self._node] = self._grad_var
+
+        def add_cand(cand):
+            if cand not in seen_set:
+                # Negate since heapq is min-heap. This is a global variable
+                heapq.heappush(cand_funcs, (-cand.rank, len(seen_set), cand))
+                seen_set.add(cand)
+
+        add_cand(self.creator_node)
+
+        while cand_funcs:
+            _, _, func = heapq.heappop(cand_funcs)
+            gxs = func.backward_accumulate(func.inputs, func.outputs, func.outputs.grad)
+
+            for x, gx in enumerate(gxs):
+                if x in grads:
+                    grads[x] += gx
+                else:
+                    grads[x] = gx
+
+                if x.creator_node is not None:
+                    add_cand(x.creator_node)
+```
+
+#### PyTorch
+Example Code
+```python
+x = Variable(torch.ones(5, 5))
+y = Variable(torch.ones(5, 5) * 4)
+z = x ** 2 + x * 2 + x * y + y
+z.backward(torch.ones(5, 5))
+```
+The trace is done by `Variable.creator` and `Function.previous_functions`.
+```python
+class Variable(object):
+    def __init__(self, tensor, creator=None, requires_grad=True):
+        if creator is None:
+            creator = Leaf(self, requires_grad)
+        self.data = tensor
+        self.creator = creator
+        self._grad = None
+
+    def backward(self, gradient=None):
+        if gradient is None:
+            if self.data.numel() != 1:
+                raise RuntimeError('backward should be called only on a scalar (i.e. 1-element tensor) or with gradient w.r.t. the variable')
+            gradient = self.data.new(1).fill_(1)
+        self._execution_engine.run_backward(self, gradient)
+
+class Function(obejct):
+    # ...
+    def _do_forward(self, *input):
+        unpacked_input = tuple(arg.data for arg in input)
+        raw_output = self.forward(*unpacked_input)
+
+        # mark output.creator = self for backward trace
+        output = tuple(Variable(tensor, self) for tensor in raw_output)
+
+        self.previous_functions = [(arg.creator, id(arg)) for arg in input]
+        self.output_ids = {id(var): i for i, var in enumerate(output)}
+        return output
+
+    def _do_backward(self, grad_output):
+        return self.backwaerd(grad_output)
+```
+The [backward](https://github.com/pytorch/pytorch/blob/v0.1.1/torch/autograd/engine.py) is similar to Autograd.
+
+#### DyNet
+Example code
+```python
+model = dy.model()
+W_p = model.add_parameters((20, 100))
+b_p = model.add_parameters(20)
+E = model.add_lookup_parameters((20000, 50))
+for epoch in range(num_epochs):
+    for in_words, out_label in training_data:
+        dy.renew_cg() # init tape
+        W = dy.parameter(W_p)
+        b = dy.parameter(b_p)
+        score_sym = dy.softmax(W*dy.concatenate([E[in_words[0]],E[in_words[1]]])+b)
+        loss_sym = dy.pickneglogsoftmax(score_sym, out_label)
+        loss_val = loss_sym.value()
+        loss_sym.backward()
+```
+[forward](https://github.com/clab/dynet/blob/740a9626a13a2732544de142e256ad0d0a166658/dynet/exec.cc#L84-L158), [backward](https://github.com/clab/dynet/blob/740a9626a13a2732544de142e256ad0d0a166658/dynet/exec.cc#L166-L284). The trace is done by creating a tape of expressions in every iteration. Backward is done by traverse the tape in the reverse order.
+```c++
+void SimpleExecutionEngine::backward(VariableIndex from_where, bool full) {
+  ...  
+  for (int i = num_nodes - 1; i >= 0; --i) {
+    // each node corresponds to an op
+    node->backward(xs, node_fx, node_dEdfx, ai, node_dEdxai);
+  }
+  ...
+}
+```

paddle/fluid/framework/CMakeLists.txt

@@ -83,8 +83,13 @@ cc_library(lod_rank_table SRCS lod_rank_table.cc DEPS lod_tensor)
 
 cc_library(feed_fetch_method SRCS feed_fetch_method.cc DEPS lod_tensor scope glog)
 
-cc_library(executor SRCS executor.cc DEPS op_registry device_context scope
-framework_proto glog lod_rank_table feed_fetch_method)
+if(WITH_DISTRIBUTE)
+  cc_library(executor SRCS executor.cc DEPS op_registry device_context scope framework_proto glog lod_rank_table feed_fetch_method sendrecvop_grpc grpc++_unsecure grpc_unsecure gpr)
+  set(DISTRIBUTE_COMPILE_FLAGS "-Wno-non-virtual-dtor -Wno-error=non-virtual-dtor -Wno-error=delete-non-virtual-dtor")
+  set_source_files_properties(executor.cc PROPERTIES COMPILE_FLAGS ${DISTRIBUTE_COMPILE_FLAGS})
+else()
+  cc_library(executor SRCS executor.cc DEPS op_registry device_context scope framework_proto glog lod_rank_table feed_fetch_method)
+endif()
 
 
 cc_library(parallel_executor SRCS parallel_executor.cc DEPS ssa_graph_builder_factory threaded_ssa_graph_executor scope_buffered_ssa_graph_executor)

paddle/fluid/framework/details/ssa_graph_checker.h

@@ -19,7 +19,7 @@
 namespace paddle {
 namespace framework {
 namespace details {
-class SSAGraph;
+struct SSAGraph;
 
 class SSAGraghBuilderWithChecker : public SSAGraphBuilder {
  public:

paddle/fluid/framework/executor.cc

@@ -20,6 +20,9 @@ limitations under the License. */
 #include "paddle/fluid/framework/lod_tensor_array.h"
 #include "paddle/fluid/framework/op_registry.h"
 #include "paddle/fluid/framework/reader.h"
+#ifdef PADDLE_WITH_DISTRIBUTE
+#include "paddle/fluid/operators/detail/grpc_client.h"
+#endif
 #include "paddle/fluid/platform/place.h"
 #include "paddle/fluid/platform/profiler.h"
 
@@ -44,6 +47,14 @@ ExecutorPrepareContext::~ExecutorPrepareContext() {
 
 Executor::Executor(const platform::Place& place) : place_(place) {}
 
+#ifdef PADDLE_WITH_DISTRIBUTE
+void Executor::Complete() {
+  ::paddle::operators::detail::RPCClient::GetInstance<
+      ::paddle::operators::detail::GRPCClient>()
+      ->SendComplete();
+}
+#endif
+
 void InitializeVariable(Variable* var, proto::VarType::Type var_type) {
   if (var_type == proto::VarType::LOD_TENSOR) {
     var->GetMutable<LoDTensor>();

paddle/fluid/framework/executor.h

@@ -44,6 +44,13 @@ class Executor {
 
   explicit Executor(const platform::Place& place);
 
+#ifdef PADDLE_WITH_DISTRIBUTE
+  /*
+   * Sending signal to pserver to mark current trainer stop.
+   */
+  void Complete();
+#endif
+
   /* @Brief
    * Runtime evaluation of the given ProgramDesc under certain Scope
    *

paddle/fluid/operators/detail/grpc_client.cc

@@ -34,6 +34,12 @@ void GRPCClient::InitEventLoop() {
   client_thread_.reset(new std::thread(std::bind(&GRPCClient::Proceed, this)));
 }
 
+void GRPCClient::SendComplete() {
+  for (auto& it : channels_) {
+    this->AsyncSendComplete(it.first);
+  }
+}
+
 GRPCClient::~GRPCClient() {
   Wait();
   cq_.Shutdown();
@@ -210,6 +216,19 @@ void GRPCClient::AsyncSendFetchBarrier(const std::string& ep,
   req_count_++;
 }
 
+void GRPCClient::AsyncSendComplete(const std::string& ep, int64_t time_out) {
+  const auto ch = GetChannel(ep);
+
+  BatchBarrierProcessor* s = new BatchBarrierProcessor(ch);
+  s->Prepare(time_out);
+
+  sendrecv::VariableMessage req;
+  req.set_varname(COMPLETE_MESSAGE);
+  auto rpc = s->stub_->AsyncSendVariable(s->context_.get(), req, &cq_);
+  rpc->Finish(&s->reply_, &s->status_, reinterpret_cast<void*>(s));
+  req_count_++;
+}
+
 void GRPCClient::Wait() {
   std::unique_lock<std::mutex> lk(sync_mutex_);
   sync_cond_.wait(lk, [this] { return req_count_ == 0; });

paddle/fluid/operators/detail/grpc_client.h

@@ -195,6 +195,8 @@ class GRPCClient : public RPCClient {
 
   void Wait() override;
 
+  void SendComplete() override;
+
  protected:
   void InitImpl() override;
 
@@ -204,6 +206,9 @@ class GRPCClient : public RPCClient {
 
   void Proceed();
 
+  void AsyncSendComplete(const std::string& ep,
+                         int64_t time_out = RPCClient::rpc_time_out);
+
   std::shared_ptr<grpc::Channel> GetChannel(const std::string& ep);
 
  private:

paddle/fluid/operators/detail/grpc_server.cc

@@ -162,16 +162,18 @@ class RequestPrefetch final : public RequestBase {
 
   void Process() override {
     // prefetch process...
-    std::string varname = request_->OutVarname();
-    VLOG(3) << "RequestPrefetch " << varname;
+    std::string in_var_name = request_->Varname();
+    std::string out_var_name = request_->OutVarname();
+    VLOG(3) << "RequestPrefetch, in_var_name: " << in_var_name
+            << " out_var_name: " << out_var_name;
 
     auto scope = request_->GetMutableLocalScope();
-    auto invar = scope->FindVar(varname);
-    framework::Variable* outvar = nullptr;
+    auto invar = scope->FindVar(in_var_name);
+    framework::Variable* outvar = scope->FindVar(out_var_name);
 
-    request_handler_->Handle(varname, scope, invar, &outvar);
+    request_handler_->Handle(in_var_name, scope, invar, &outvar, out_var_name);
 
-    SerializeToByteBuffer(varname, outvar, *request_handler_->dev_ctx(),
+    SerializeToByteBuffer(out_var_name, outvar, *request_handler_->dev_ctx(),
                           &reply_);
     Finish(reply_, &responder_);
   }
@@ -287,7 +289,7 @@ void AsyncGRPCServer::TryToRegisterNewOne(const std::string& rpc_name,
   } else if (rpc_name == kRequestPrefetch) {
     b = new RequestPrefetch(&service_, cq.get(), handler, req_id);
   } else {
-    PADDLE_ENFORCE(false, "not surpported rpc");
+    PADDLE_ENFORCE(false, "not supported rpc");
   }
 
   reqs[req_id] = b;

paddle/fluid/operators/detail/request_handler.h

@@ -40,6 +40,7 @@ constexpr char kRequestPrefetch[] = "RequestPrefetch";
 #define LISTEN_TERMINATE_MESSAGE "TERMINATE@RECV"
 #define BATCH_BARRIER_MESSAGE "BATCH_BARRIER@RECV"
 #define FETCH_BARRIER_MESSAGE "FETCH_BARRIER@RECV"
+#define COMPLETE_MESSAGE "COMPLETE@RECV"
 
 class RPCServer;
 
@@ -60,9 +61,12 @@ class RequestHandler {
   void SetDevCtx(const platform::DeviceContext* dev_ctx) { dev_ctx_ = dev_ctx; }
   void SetProgram(framework::ProgramDesc* program) { program_ = program; }
   void SetExecutor(framework::Executor* executor) { executor_ = executor; }
+
+  // Used for dist lookup table prefetch
   void SetPrefetchPreparedCtx(
-      std::unique_ptr<framework::ExecutorPrepareContext> prepared) {
-    prefetch_ctx_.reset(prepared.release());
+      std::unordered_map<
+          std::string, std::shared_ptr<framework::ExecutorPrepareContext>>* g) {
+    prefetch_var_name_to_prepared_ctx_ = g;
   }
 
   // Used for async.
@@ -78,9 +82,6 @@ class RequestHandler {
   bool sync_mode() { return sync_mode_; }
   framework::Scope* scope() { return scope_; }
   const platform::DeviceContext* dev_ctx() { return dev_ctx_; }
-  framework::ExecutorPrepareContext* prefetch_ctx() {
-    return prefetch_ctx_.get();
-  }
   framework::ProgramDesc* program() { return program_; }
   framework::Executor* executor() { return executor_; }
 
@@ -99,8 +100,8 @@ class RequestHandler {
   //           *request_handler_->dev_ctx(), &reply_);
   //    }
   virtual bool Handle(const std::string& varname, framework::Scope* scope,
-                      framework::Variable* var,
-                      framework::Variable** outvar) = 0;
+                      framework::Variable* var, framework::Variable** outvar,
+                      const std::string& out_var_name = "") = 0;
 
  protected:
   const bool sync_mode_;
@@ -109,12 +110,17 @@ class RequestHandler {
   framework::Executor* executor_;
   framework::Scope* scope_;
   framework::ProgramDesc* program_;
-  std::unique_ptr<framework::ExecutorPrepareContext> prefetch_ctx_;
+
+  // used for distribute lookup table prefetch
+  std::unordered_map<std::string,
+                     std::shared_ptr<framework::ExecutorPrepareContext>>*
+      prefetch_var_name_to_prepared_ctx_;
 
   // Used for async.
   std::unordered_map<std::string,
                      std::shared_ptr<framework::ExecutorPrepareContext>>*
       grad_to_prepared_ctx_;
+
   RPCServer* rpc_server_;
 };
 

paddle/fluid/operators/detail/request_handler_impl.cc

@@ -30,7 +30,8 @@ namespace detail {
 bool RequestSendHandler::Handle(const std::string& varname,
                                 framework::Scope* scope,
                                 framework::Variable* invar,
-                                framework::Variable** outvar) {
+                                framework::Variable** outvar,
+                                const std::string& out_var_name) {
   VLOG(4) << "RequestSendHandler:" << varname;
 
   // Async
@@ -49,6 +50,9 @@ bool RequestSendHandler::Handle(const std::string& varname,
   if (varname == BATCH_BARRIER_MESSAGE) {
     VLOG(3) << "sync: recv batch barrier message";
     rpc_server_->IncreaseBatchBarrier(kRequestSend);
+  } else if (varname == COMPLETE_MESSAGE) {
+    VLOG(3) << "sync: recv complete message";
+    rpc_server_->DecreaseClientNum();
   } else {
     VLOG(3) << "sync: received var_name: " << varname;
     if (sync_mode_) {
@@ -79,7 +83,8 @@ void RequestSendHandler::ResetSparseVarRecorder() {
 bool RequestGetHandler::Handle(const std::string& varname,
                                framework::Scope* scope,
                                framework::Variable* invar,
-                               framework::Variable** outvar) {
+                               framework::Variable** outvar,
+                               const std::string& out_var_name) {
   VLOG(4) << "RequestGetHandler:" << varname;
 
   if (varname != FETCH_BARRIER_MESSAGE) {
@@ -102,13 +107,14 @@ bool RequestGetHandler::Handle(const std::string& varname,
 bool RequestPrefetchHandler::Handle(const std::string& varname,
                                     framework::Scope* scope,
                                     framework::Variable* invar,
-                                    framework::Variable** outvar) {
+                                    framework::Variable** outvar,
+                                    const std::string& out_var_name) {
   VLOG(4) << "RequestPrefetchHandler " << varname;
 
-  auto var_desc = program_->Block(0).FindVar(varname);
-  *outvar = scope->FindVar(varname);
+  auto var_desc = program_->Block(0).FindVar(out_var_name);
   InitializeVariable(*outvar, var_desc->GetType());
-  executor_->RunPreparedContext(prefetch_ctx_.get(), scope);
+  executor_->RunPreparedContext(
+      (*prefetch_var_name_to_prepared_ctx_)[varname].get(), scope);
 
   return true;
 }

paddle/fluid/operators/detail/request_handler_impl.h

@@ -39,7 +39,8 @@ class RequestSendHandler final : public RequestHandler {
   explicit RequestSendHandler(bool sync_mode) : RequestHandler(sync_mode) {}
   virtual ~RequestSendHandler() {}
   bool Handle(const std::string& varname, framework::Scope* scope,
-              framework::Variable* var, framework::Variable** outvar) override;
+              framework::Variable* var, framework::Variable** outvar,
+              const std::string& out_var_name = "") override;
   void ResetSparseVarRecorder();
 
  private:
@@ -52,7 +53,8 @@ class RequestGetHandler final : public RequestHandler {
   explicit RequestGetHandler(bool sync_mode) : RequestHandler(sync_mode) {}
   virtual ~RequestGetHandler() {}
   bool Handle(const std::string& varname, framework::Scope* scope,
-              framework::Variable* var, framework::Variable** outvar) override;
+              framework::Variable* var, framework::Variable** outvar,
+              const std::string& out_var_name = "") override;
 };
 
 class RequestPrefetchHandler final : public RequestHandler {
@@ -60,7 +62,8 @@ class RequestPrefetchHandler final : public RequestHandler {
   explicit RequestPrefetchHandler(bool sync_mode) : RequestHandler(sync_mode) {}
   virtual ~RequestPrefetchHandler() {}
   bool Handle(const std::string& varname, framework::Scope* scope,
-              framework::Variable* var, framework::Variable** outvar) override;
+              framework::Variable* var, framework::Variable** outvar,
+              const std::string& out_var_name = "") override;
 };
 
 }  // namespace detail

paddle/fluid/operators/detail/rpc_client.h

@@ -53,6 +53,11 @@ class RPCClient {
   virtual void AsyncSendFetchBarrier(const std::string& ep,
                                      int64_t time_out = rpc_time_out) = 0;
 
+  // SendComplete tells all the server that current trainer have no more data
+  // to train, so that the pserver can reduce it's barrier count, and continue
+  // to train with other trainers.
+  virtual void SendComplete() = 0;
+
   virtual void Wait() = 0;
 
   static constexpr int64_t rpc_time_out = 120 * 1000;

paddle/fluid/operators/detail/rpc_server.cc

@@ -43,7 +43,7 @@ void RPCServer::SavePort() const {
 
 void RPCServer::WaitBarrier(const std::string& rpc_name) {
   std::unique_lock<std::mutex> lock(this->mutex_);
-  barrier_cond_.wait(lock, [=] {
+  barrier_cond_.wait(lock, [this, &rpc_name] {
     return (barrier_counter_[rpc_name] >= client_num_ || exit_flag_.load());
   });
 
@@ -53,19 +53,23 @@ void RPCServer::WaitBarrier(const std::string& rpc_name) {
 void RPCServer::IncreaseBatchBarrier(const std::string rpc_name) {
   VLOG(3) << "RPCServer begin IncreaseBatchBarrier " << rpc_name;
   int b = 0;
-  {
-    std::unique_lock<std::mutex> lock(mutex_);
-    b = ++barrier_counter_[rpc_name];
-  }
-
-  VLOG(3) << "RPCServer IncreaseBatchBarrier " << rpc_name
-          << ", barrier_count:" << b << ", fan_in" << client_num_;
-
+  std::unique_lock<std::mutex> lock(mutex_);
+  b = ++barrier_counter_[rpc_name];
   if (b >= client_num_) {
+    lock.unlock();
     barrier_cond_.notify_all();
+    lock.lock();
   }
 }
 
+void RPCServer::DecreaseClientNum() {
+  {
+    std::unique_lock<std::mutex> lock(mutex_);
+    client_num_--;
+  }
+  barrier_cond_.notify_all();
+}
+
 void RPCServer::ResetBarrierCounter() {
   VLOG(3) << "RPCServer ResetBarrierCounter ";
   std::unique_lock<std::mutex> lock(mutex_);

paddle/fluid/operators/detail/rpc_server.h

@@ -60,7 +60,7 @@ class RPCServer {
   void SetCond(const std::string& rpc_name);
   void WaitCond(const std::string& rpc_name);
   void IncreaseBatchBarrier(const std::string rpc_name);
-
+  void DecreaseClientNum();
   void ResetBarrierCounter();
 
  protected:
@@ -79,8 +79,7 @@ class RPCServer {
   std::string bind_address_;
   std::atomic<int> exit_flag_;
   int selected_port_;
-
-  const int client_num_;
+  int client_num_;
 
   std::unordered_map<std::string, RequestHandler*> rpc_call_map_;
   std::unordered_map<std::string, int> rpc_thread_num_;

paddle/fluid/operators/detail/rpc_server_test.cc

@@ -98,11 +98,17 @@ void StartServer() {
   framework::Executor exe(place);
   platform::CPUDeviceContext ctx(place);
   auto* block = AppendPrefetchBlcok(&program);
-  auto prepared = exe.Prepare(program, block->ID());
+  std::string in_var_name("ids");
+  std::vector<int> prefetch_block_ids{block->ID()};
+  auto prepared = exe.Prepare(program, prefetch_block_ids);
   InitTensorsOnServer(&scope, &place, 10);
 
+  std::unordered_map<std::string,
+                     std::shared_ptr<framework::ExecutorPrepareContext>>
+      prefetch_var_name_to_prepared;
+  prefetch_var_name_to_prepared[in_var_name] = prepared[0];
   g_req_handler->SetProgram(&program);
-  g_req_handler->SetPrefetchPreparedCtx(std::move(prepared));
+  g_req_handler->SetPrefetchPreparedCtx(&prefetch_var_name_to_prepared);
   g_req_handler->SetDevCtx(&ctx);
   g_req_handler->SetScope(&scope);
   g_req_handler->SetExecutor(&exe);

paddle/fluid/operators/elementwise_op.h

@@ -66,40 +66,41 @@ class ElementwiseOpMaker : public framework::OpProtoAndCheckerMaker {
         .SetDefault(-1)
         .EqualGreaterThan(-1);
     AddComment(string::Sprintf(R"DOC(
-Limited Elementwise %s Operator.
+Limited Elementwise %s Operator
 
 The equation is:
 
 $$%s$$
 
-$X$ is a tensor of any dimension and the dimensions of tensor $Y$ must be
-smaller than or equal to the dimensions of $X$.
+- $X$: a tensor of any dimension. 
+- $Y$: a tensor whose dimensions must be less than or equal to the dimensions of $X$.
 
 There are two cases for this operator:
-1. The shape of $Y$ is same with $X$;
-2. The shape of $Y$ is a congiguous subsequencet of $X$. The trailing dimensions
-   of size 1 for $Y$ will be ignored for the consideration of subsequence.
 
+1. The shape of $Y$ is the same with $X$.
+2. The shape of $Y$ is a continuous subsequence of $X$.
 
 For case 2:
 
-$Y$ will be broadcasted to match the shape of $X$ and axis should be
-set to index of the start dimension to broadcast $Y$ onto $X$.
+1. Broadcast $Y$ to match the shape of $X$, where $axis$ is the start dimension index 
+   for broadcasting $Y$ onto $X$. 
+2. If $axis$ is -1 (default), $axis = rank(X) - rank(Y)$.
+3. The trailing dimensions of size 1 for $Y$ will be ignored for the consideration of 
+   subsequence, such as shape(Y) = (2, 1) => (2).
 
-If axis is -1, it is treated as axis=rank(X)-rank(Y).
+For example:
 
-For example
   .. code-block:: python
 
     shape(X) = (2, 3, 4, 5), shape(Y) = (,)
     shape(X) = (2, 3, 4, 5), shape(Y) = (5,)
-    shape(X) = (2, 3, 4, 5), shape(Y) = (4, 5)
+    shape(X) = (2, 3, 4, 5), shape(Y) = (4, 5), with axis=-1(default) or axis=2
     shape(X) = (2, 3, 4, 5), shape(Y) = (3, 4), with axis=1
     shape(X) = (2, 3, 4, 5), shape(Y) = (2), with axis=0
     shape(X) = (2, 3, 4, 5), shape(Y) = (2, 1), with axis=0
 
-Either of the inputs $X$ and $Y$ or none can carry the LoD (Level of Details)
-information. However, the output only shares the LoD information with input $X$.
+The inputs $X$ and $Y$ can carry the different LoD information. 
+But the output only shares the LoD information with the input $X$.
 
 )DOC",
                                GetName(), GetEquation()));

paddle/fluid/operators/listen_and_serv_op.cc

@@ -96,19 +96,22 @@ static int64_t GetTimestamp() {
   return tp.tv_sec * 1000 + tp.tv_usec / 1000;
 }
 
-void ListenAndServOp::RunSyncLoop(framework::Executor *executor,
-                                  framework::ProgramDesc *program,
-                                  framework::Scope *recv_scope,
-                                  framework::BlockDesc *prefetch_block) const {
+void ListenAndServOp::RunSyncLoop(
+    framework::Executor *executor, framework::ProgramDesc *program,
+    framework::Scope *recv_scope,
+    const std::vector<int> &prefetch_block_id_list) const {
   size_t num_blocks = program->Size();
   PADDLE_ENFORCE_GE(num_blocks, 2,
                     "server program should have at least 2 blocks");
 
-  std::vector<int> block_list;
-  for (size_t blkid = 1; blkid < num_blocks; ++blkid) {
-    block_list.push_back(blkid);
+  std::vector<int> optimize_block_id_list;
+  for (int blkid = 1; blkid < num_blocks; ++blkid) {
+    if (std::find(prefetch_block_id_list.begin(), prefetch_block_id_list.end(),
+                  blkid) == prefetch_block_id_list.end()) {
+      optimize_block_id_list.push_back(blkid);
+    }
   }
-  auto optimize_prepared = executor->Prepare(*program, block_list);
+  auto optimize_prepared = executor->Prepare(*program, optimize_block_id_list);
   // Insert placeholder for block0 which holds current op itself.
   optimize_prepared.insert(
       optimize_prepared.begin(),
@@ -135,16 +138,17 @@ void ListenAndServOp::RunSyncLoop(framework::Executor *executor,
     std::vector<size_t> parallel_blkids;
     parallel_blkids.push_back(1);
     double ts = GetTimestamp();
-    for (size_t blkid = 2; blkid < num_blocks; ++blkid) {
-      if (blkid != static_cast<size_t>(prefetch_block->ID())) {
-        if (program->Block(blkid).Parent() != last_parent_blkid) {
-          ParallelExecuteBlocks(parallel_blkids, executor, optimize_prepared,
-                                program, recv_scope);
-          parallel_blkids.clear();
-          last_parent_blkid = program->Block(blkid).Parent();
-        }
-        parallel_blkids.push_back(blkid);
+    for (size_t i = 1; i < optimize_block_id_list.size(); ++i) {
+      // skip the first optimize block because it is already in the
+      // parallel_blkids.
+      int blkid = optimize_block_id_list[i];
+      if (program->Block(blkid).Parent() != last_parent_blkid) {
+        ParallelExecuteBlocks(parallel_blkids, executor, optimize_prepared,
+                              program, recv_scope);
+        parallel_blkids.clear();
+        last_parent_blkid = program->Block(blkid).Parent();
       }
+      parallel_blkids.push_back(blkid);
     }
     ParallelExecuteBlocks(parallel_blkids, executor, optimize_prepared, program,
                           recv_scope);
@@ -210,18 +214,19 @@ void ListenAndServOp::RunAsyncLoop(framework::Executor *executor,
   }  // while(true)
 }
 
-static void FillRequestCtx(detail::RequestHandler *h, framework::Scope *scope,
-                           platform::DeviceContext *dev_ctx,
-                           framework::Executor *executor,
-                           framework::ProgramDesc *program,
-                           framework::ExecutorPrepareContext *prefetch_ctx,
-                           detail::RPCServer *rpc_server) {
+static void FillRequestCtx(
+    detail::RequestHandler *h, framework::Scope *scope,
+    platform::DeviceContext *dev_ctx, framework::Executor *executor,
+    framework::ProgramDesc *program,
+    std::unordered_map<std::string,
+                       std::shared_ptr<framework::ExecutorPrepareContext>>
+        *prefetch_ctx,
+    detail::RPCServer *rpc_server) {
   h->SetScope(scope);
   h->SetDevCtx(dev_ctx);
   h->SetExecutor(executor);
   h->SetProgram(program);
-  h->SetPrefetchPreparedCtx(
-      std::unique_ptr<framework::ExecutorPrepareContext>(prefetch_ctx));
+  h->SetPrefetchPreparedCtx(prefetch_ctx);
   h->SetRPCServer(rpc_server);
 }
 
@@ -255,17 +260,42 @@ void ListenAndServOp::RunImpl(const framework::Scope &scope,
                             request_prefetch_handler_.get());
 
   auto *optimize_block = Attr<framework::BlockDesc *>(kOptimizeBlock);
-  auto *prefetch_block = Attr<framework::BlockDesc *>(kPrefetchBlock);
   auto *program = optimize_block->Program();
   framework::Executor executor(dev_place);
 
   // prepare for prefetch
-  VLOG(3) << "prefetch block id is " << prefetch_block->ID();
-  auto prefetch_prepared = executor.Prepare(*program, prefetch_block->ID());
+  std::vector<int> prefetch_block_id_list;
+  std::unordered_map<int, std::string> block_id_to_prefetch_var_name;
+
+  auto prefetch_var_name_to_block_id_str =
+      Attr<std::vector<std::string>>(kPrefetchVarNameToBlockId);
+  for (const auto &prefetch_var_name_and_id :
+       prefetch_var_name_to_block_id_str) {
+    std::vector<std::string> pieces;
+    split(prefetch_var_name_and_id, ':', &pieces);
+    VLOG(3) << "after split, prefetch_var = " << pieces[0]
+            << ", id=" << pieces[1];
+    PADDLE_ENFORCE_EQ(pieces.size(), 2);
+
+    int block_id = std::stoi(pieces[1]);
+    prefetch_block_id_list.push_back(block_id);
+    block_id_to_prefetch_var_name[block_id] = pieces[0];
+  }
+
+  auto prefetch_prepared = executor.Prepare(*program, prefetch_block_id_list);
+
+  std::unordered_map<std::string,
+                     std::shared_ptr<framework::ExecutorPrepareContext>>
+      prefetch_var_name_to_prepared_ctx;
+  for (size_t i = 0; i < prefetch_block_id_list.size(); ++i) {
+    auto block_id = prefetch_block_id_list[i];
+    auto prefetch_var_name = block_id_to_prefetch_var_name[block_id];
+    prefetch_var_name_to_prepared_ctx[prefetch_var_name] = prefetch_prepared[i];
+  }
 
   auto f = std::bind(FillRequestCtx, std::placeholders::_1, &recv_scope,
-                     &dev_ctx, &executor, program, prefetch_prepared.release(),
-                     rpc_service_.get());
+                     &dev_ctx, &executor, program,
+                     &prefetch_var_name_to_prepared_ctx, rpc_service_.get());
 
   f(request_send_handler_.get());
   f(request_get_handler_.get());
@@ -283,7 +313,7 @@ void ListenAndServOp::RunImpl(const framework::Scope &scope,
   // Write to a file of server selected port for python use.
   SavePort();
   if (sync_mode) {
-    RunSyncLoop(&executor, program, &recv_scope, prefetch_block);
+    RunSyncLoop(&executor, program, &recv_scope, prefetch_block_id_list);
   } else {
     RunAsyncLoop(&executor, program);
   }
@@ -309,8 +339,9 @@ class ListenAndServOpMaker : public framework::OpProtoAndCheckerMaker {
     AddAttr<bool>("sync_mode", "if works at sync_mode or not").SetDefault(true);
     AddAttr<framework::BlockDesc *>(kOptimizeBlock,
                                     "BlockID to run on server side.");
-    AddAttr<framework::BlockDesc *>(kPrefetchBlock,
-                                    "prefetch block to run on server side.");
+    AddAttr<std::vector<std::string>>(kPrefetchVarNameToBlockId,
+                                      "prefetch blocks to run on server side.")
+        .SetDefault({});
     AddAttr<int>("Fanin", "How many clients send to this server.")
         .SetDefault(1);
   }

paddle/fluid/operators/listen_and_serv_op.h

@@ -18,6 +18,7 @@ limitations under the License. */
 #include <atomic>
 #include <set>
 #include <string>
+#include <vector>
 
 #include "paddle/fluid/framework/executor.h"
 #include "paddle/fluid/framework/lod_tensor.h"
@@ -30,7 +31,7 @@ namespace paddle {
 namespace operators {
 
 constexpr char kOptimizeBlock[] = "OptimizeBlock";
-constexpr char kPrefetchBlock[] = "PrefetchBlock";
+constexpr char kPrefetchVarNameToBlockId[] = "prefetch_var_name_to_block_id";
 
 void RunServer(std::shared_ptr<detail::RPCServer> service);
 
@@ -46,7 +47,7 @@ class ListenAndServOp : public framework::OperatorBase {
   void RunSyncLoop(framework::Executor* executor,
                    framework::ProgramDesc* program,
                    framework::Scope* recv_scope,
-                   framework::BlockDesc* prefetch_block) const;
+                   const std::vector<int>& prefetch_block_id_list) const;
 
   void RunAsyncLoop(framework::Executor* executor,
                     framework::ProgramDesc* program) const;

paddle/fluid/operators/norm_op.cc

@@ -16,40 +16,34 @@ limitations under the License. */
 namespace paddle {
 namespace operators {
 
-template <typename AttrType>
 class NormOpMaker : public framework::OpProtoAndCheckerMaker {
  public:
   void Make() override {
-    AddInput(
-        "X",
-        "(Tensor) The input tensor of norm operator. "
-        "The format of input tensor is NCHW. Where N is batch size, C is the "
-        "number of channels, H and W is the height and width of feature.");
-    AddInput("Scale",
-             "(Tensor) The input tensor of norm operator. "
-             "The format of input tensor is C * 1.");
-    AddAttr<AttrType>("epsilon",
-                      "(float, default 1e-10) Constant "
-                      "for numerical stability.")
+    AddInput("X", "(Tensor) A tensor of rank >= axis.");
+    AddAttr<int>("axis",
+                 "The axis on which to apply normalization. If axis < 0, "
+                 "the dimension to normalization is rank(X) + axis. -1 is "
+                 "the last dimension.");
+    AddAttr<float>("epsilon",
+                   "(float, default 1e-10) The epsilon value is used "
+                   "to avoid division by zero.")
         .SetDefault(1.0e-10f);
-    AddOutput("Out",
-              "(Tensor) The output tensor of norm operator."
-              "N * M."
-              "M = C * H * W");
+    AddOutput("Norm",
+              "(Tensor) A tensor saved the `sqrt(sum(x) + epsion)` will "
+              "be used in backward kernel.")
+        .AsIntermediate();
+    AddOutput("Out", "(Tensor) A tensor of the same shape as X.");
     AddComment(R"DOC(
-       "Input shape: $(N, C, H, W)$
-        Scale shape: $(C, 1)$
-        Output shape: $(N, C, H, W)$
-        Where
-        forward
-          $$
-            [\frac {x_{1}}{\sqrt{\sum{x_{i}^{2}}}} \frac {x_{2}}{\sqrt{\sum{x_{i}^{2}}}} \frac {x_{3}}{\sqrt{\sum{x_{i}^{2}}}} \cdot  \cdot  \cdot \frac {x_{n}}{\sqrt{\sum{x_{i}^{2}}}}]
-          $$
-        backward
-          $$
-            \frac{\frac{\mathrm{d}L }{\mathrm{d}y_{1}} - \frac {x_{1}\sum {\frac{\mathrm{d} L}{\mathrm{d} y_{j}}}x_{j}}{\sum x_{j}^{2}} }{\sqrt{\sum{x_{j}^{2}}}}
-          $$
-        )DOC");
+
+Given a tensor, apply 2-normalization along the provided axis.
+
+$$
+y = \frac{x}{ \sqrt{\sum {x^2} + epsion }}
+$$
+
+where, $\sum {x^2}$ is calculated along the `axis` dimension.
+        
+)DOC");
   }
 };
 
@@ -58,15 +52,15 @@ class NormOp : public framework::OperatorWithKernel {
   using framework::OperatorWithKernel::OperatorWithKernel;
   void InferShape(framework::InferShapeContext* ctx) const override {
     PADDLE_ENFORCE(ctx->HasInput("X"),
-                   "Input(X) of NormOp"
-                   "should not be null.");
-    PADDLE_ENFORCE(ctx->HasInput("Scale"),
-                   "Input(Scale) of NormOp"
-                   "should not be null.");
+                   "Input(X) of NormOp should not be null.");
     PADDLE_ENFORCE(ctx->HasOutput("Out"),
                    "Output(Out) of NormOp should not be null.");
-    auto in_x_dims = ctx->GetInputDim("X");
-    ctx->SetOutputDim("Out", in_x_dims);
+    auto xdim = ctx->GetInputDim("X");
+    ctx->SetOutputDim("Out", xdim);
+    int axis = ctx->Attrs().Get<int>("axis");
+    if (axis < 0) axis = xdim.size() + axis;
+    xdim[axis] = 1;
+    ctx->SetOutputDim("Norm", xdim);
   }
 };
 
@@ -84,12 +78,12 @@ class NormOpGrad : public framework::OperatorWithKernel {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OPERATOR(norm, ops::NormOp, ops::NormOpMaker<float>,
+using CPU = paddle::platform::CPUDeviceContext;
+
+REGISTER_OPERATOR(norm, ops::NormOp, ops::NormOpMaker,
                   paddle::framework::DefaultGradOpDescMaker<true>);
 REGISTER_OPERATOR(norm_grad, ops::NormOpGrad);
-REGISTER_OP_CPU_KERNEL(
-    norm, ops::NormKernel<paddle::platform::CPUDeviceContext, float>,
-    ops::NormKernel<paddle::platform::CPUDeviceContext, double, float>);
-REGISTER_OP_CPU_KERNEL(
-    norm_grad, ops::NormGradKernel<paddle::platform::CPUDeviceContext, float>,
-    ops::NormGradKernel<paddle::platform::CPUDeviceContext, double, float>);
+REGISTER_OP_CPU_KERNEL(norm, ops::NormKernel<CPU, float>,
+                       ops::NormKernel<CPU, double>);
+REGISTER_OP_CPU_KERNEL(norm_grad, ops::NormGradKernel<CPU, float>,
+                       ops::NormGradKernel<CPU, double>);

paddle/fluid/operators/norm_op.cu

@@ -16,9 +16,9 @@ limitations under the License. */
 #include "paddle/fluid/operators/norm_op.h"
 
 namespace ops = paddle::operators;
-REGISTER_OP_CUDA_KERNEL(
-    norm, ops::NormKernel<paddle::platform::CUDADeviceContext, float>,
-    ops::NormKernel<paddle::platform::CUDADeviceContext, double, float>);
-REGISTER_OP_CUDA_KERNEL(
-    norm_grad, ops::NormGradKernel<paddle::platform::CUDADeviceContext, float>,
-    ops::NormGradKernel<paddle::platform::CUDADeviceContext, double, float>);
+using CUDA = paddle::platform::CUDADeviceContext;
+
+REGISTER_OP_CUDA_KERNEL(norm, ops::NormKernel<CUDA, float>,
+                        ops::NormKernel<CUDA, double>);
+REGISTER_OP_CUDA_KERNEL(norm_grad, ops::NormGradKernel<CUDA, float>,
+                        ops::NormGradKernel<CUDA, double>);

paddle/fluid/operators/norm_op.h

@@ -19,156 +19,110 @@ limitations under the License. */
 namespace paddle {
 namespace operators {
 
-template <typename DeviceContext, typename T, typename AttrType = T>
+inline void GetDims(const framework::DDim& dim, int axis, int* pre, int* n,
+                    int* post) {
+  *pre = 1;
+  *post = 1;
+  *n = dim[axis];
+  for (int i = 0; i < axis; ++i) {
+    (*pre) *= dim[i];
+  }
+  for (int i = axis + 1; i < dim.size(); ++i) {
+    (*post) *= dim[i];
+  }
+}
+
+template <typename DeviceContext, typename T>
 class NormKernel : public framework::OpKernel<T> {
  public:
-  void Compute(const framework::ExecutionContext& context) const override {
-    const framework::Tensor* in_x = context.Input<framework::Tensor>("X");
-    const framework::Tensor* scale = context.Input<framework::Tensor>("Scale");
-    auto* out = context.Output<framework::Tensor>("Out");
-    auto epsilon = static_cast<T>(context.Attr<AttrType>("epsilon"));
-    out->mutable_data<T>(context.GetPlace());
-    int batch_size = in_x->dims()[0];
-    int channels = in_x->dims()[1];
-    int height = in_x->dims()[2];
-    int width = in_x->dims()[3];
-    int fea_len = height * width;
-    auto* place =
-        context.template device_context<DeviceContext>().eigen_device();
-    auto x =
-        framework::EigenMatrix<T, Eigen::RowMajor, Eigen::DenseIndex>::From(
-            *in_x, framework::make_ddim({batch_size, fea_len * channels}));
-    // get square
-    framework::Tensor x_square;
-    x_square.mutable_data<T>(in_x->dims(), context.GetPlace());
-    auto x_square_eigen =
-        framework::EigenMatrix<T, Eigen::RowMajor, Eigen::DenseIndex>::From(
-            x_square, framework::make_ddim({batch_size, fea_len * channels}));
-    x_square_eigen.device(*place) = x.square();
-    auto scale_eigen =
-        framework::EigenVector<T, Eigen::RowMajor, Eigen::DenseIndex>::Flatten(
-            *scale);
-    for (int n = 0; n < batch_size; ++n) {
-      framework::Tensor in_x_batch = in_x->Slice(n, n + 1);
-      auto in_x_batch_eigen =
-          framework::EigenMatrix<T, Eigen::RowMajor, Eigen::DenseIndex>::From(
-              in_x_batch, framework::make_ddim({channels, fea_len}));
-      framework::Tensor x_square_batch = x_square.Slice(n, n + 1);
-      auto x_square_batch_eigen =
-          framework::EigenMatrix<T, Eigen::RowMajor, Eigen::DenseIndex>::From(
-              x_square_batch, framework::make_ddim({channels, fea_len}));
-      framework::Tensor out_batch = out->Slice(n, n + 1);
-      auto out_batch_eigen =
-          framework::EigenMatrix<T, Eigen::RowMajor, Eigen::DenseIndex>::From(
-              out_batch, framework::make_ddim({channels, fea_len}));
-      framework::Tensor tmp_tensor;
-      tmp_tensor.mutable_data<T>(framework::make_ddim({1, fea_len}),
-                                 context.GetPlace());
-      auto tmp = framework::EigenVector<T, Eigen::RowMajor,
-                                        Eigen::DenseIndex>::Flatten(tmp_tensor);
-      // get colsum and sqrt , inverse
-      auto dim = Eigen::array<int, 1>({{0}});
-      tmp.device(*place) = x_square_batch_eigen.sum(dim);
-      tmp.device(*place) = (tmp + epsilon).sqrt().inverse();
-      Eigen::array<int, 2> broadcast_dim_col;
-      broadcast_dim_col[1] = 1;
-      broadcast_dim_col[0] = channels;
-      out_batch_eigen.device(*place) =
-          in_x_batch_eigen * (tmp.broadcast(broadcast_dim_col));
-      Eigen::array<int, 2> broadcast_dim_row;
-      broadcast_dim_row[1] = fea_len;
-      broadcast_dim_row[0] = 1;
-      out_batch_eigen.device(*place) =
-          out_batch_eigen * (scale_eigen.broadcast(broadcast_dim_row));
-    }
+  void Compute(const framework::ExecutionContext& ctx) const override {
+    auto* in_x = ctx.Input<framework::Tensor>("X");
+    auto* out_y = ctx.Output<framework::Tensor>("Out");
+    auto* out_norm = ctx.Output<framework::Tensor>("Norm");
+    out_y->mutable_data<T>(ctx.GetPlace());
+    out_norm->mutable_data<T>(ctx.GetPlace());
+
+    auto xdim = in_x->dims();
+    auto ndim = out_norm->dims();
+    T eps = static_cast<T>(ctx.Attr<float>("epsilon"));
+    int axis = ctx.Attr<int>("axis");
+    if (axis < 0) axis = xdim.size() + axis;
+    int pre, n, post;
+    GetDims(xdim, axis, &pre, &n, &post);
+
+    auto* place = ctx.template device_context<DeviceContext>().eigen_device();
+
+    Eigen::DSizes<int, 3> shape(pre, n, post);
+    Eigen::DSizes<int, 2> norm_shape(pre, post);
+
+    auto x_e = framework::EigenVector<T>::Flatten(*in_x);
+    auto y_e = framework::EigenVector<T>::Flatten(*out_y);
+    auto norm_e = framework::EigenVector<T>::Flatten(*out_norm);
+    auto x = x_e.reshape(shape);
+    auto y = y_e.reshape(shape);
+    auto norm = norm_e.reshape(norm_shape);
+
+    Eigen::DSizes<int, 1> rdim(1);
+    // y = x / sqrt((sum(x * x) + epsilon))
+    // norm = sqrt(sum(x * x) + epsilon)
+    auto sum = x.pow(2).sum(rdim) + eps;
+    norm.device(*place) = sum.sqrt();
+    // y = x / norm
+    Eigen::DSizes<int, 3> rshape(pre, 1, post);
+    Eigen::DSizes<int, 3> bcast(1, n, 1);
+    y.device(*place) = x / norm.reshape(rshape).broadcast(bcast);
   }
 };
 template <typename DeviceContext, typename T, typename AttrType = T>
 class NormGradKernel : public framework::OpKernel<T> {
  public:
-  void Compute(const framework::ExecutionContext& context) const override {
-    const framework::Tensor* in_x = context.Input<framework::Tensor>("X");
-    const framework::Tensor* scale = context.Input<framework::Tensor>("Scale");
-    const framework::Tensor* out_grad =
-        context.Input<framework::Tensor>(framework::GradVarName("Out"));
-    auto epsilon = static_cast<T>(context.Attr<AttrType>("epsilon"));
-    framework::Tensor* in_x_grad =
-        context.Output<framework::Tensor>(framework::GradVarName("X"));
-    in_x_grad->mutable_data<T>(context.GetPlace());
-    int batch_size = in_x->dims()[0];
-    int channels = in_x->dims()[1];
-    int height = in_x->dims()[2];
-    int width = in_x->dims()[3];
-    int fea_len = height * width;
-    auto* place =
-        context.template device_context<DeviceContext>().eigen_device();
-
-    auto scale_eigen =
-        framework::EigenVector<T, Eigen::RowMajor, Eigen::DenseIndex>::Flatten(
-            *scale);
-    auto x =
-        framework::EigenMatrix<T, Eigen::RowMajor, Eigen::DenseIndex>::From(
-            *in_x, framework::make_ddim({batch_size, fea_len * channels}));
-    // get square
-    framework::Tensor x_square;
-    x_square.mutable_data<T>(in_x->dims(), context.GetPlace());
-    auto x_square_eigen =
-        framework::EigenMatrix<T, Eigen::RowMajor, Eigen::DenseIndex>::From(
-            x_square, framework::make_ddim({batch_size, fea_len * channels}));
-    x_square_eigen.device(*place) = x.square();
-
-    for (int n = 0; n < batch_size; ++n) {
-      framework::Tensor in_x_batch = in_x->Slice(n, n + 1);
-      auto in_x_batch_eigen =
-          framework::EigenMatrix<T, Eigen::RowMajor, Eigen::DenseIndex>::From(
-              in_x_batch, framework::make_ddim({channels, fea_len}));
-      framework::Tensor in_g_batch = in_x_grad->Slice(n, n + 1);
-      auto in_g_batch_eigen =
-          framework::EigenMatrix<T, Eigen::RowMajor, Eigen::DenseIndex>::From(
-              in_g_batch, framework::make_ddim({channels, fea_len}));
-      framework::Tensor x_square_batch = x_square.Slice(n, n + 1);
-      auto x_square_batch_eigen =
-          framework::EigenMatrix<T, Eigen::RowMajor, Eigen::DenseIndex>::From(
-              x_square_batch, framework::make_ddim({channels, fea_len}));
-      framework::Tensor outg_batch = out_grad->Slice(n, n + 1);
-      auto outg_batch_eigen =
-          framework::EigenMatrix<T, Eigen::RowMajor, Eigen::DenseIndex>::From(
-              outg_batch, framework::make_ddim({channels, fea_len}));
-
-      framework::Tensor tmp_tensor;
-      tmp_tensor.mutable_data<T>(framework::make_ddim({1, fea_len}),
-                                 context.GetPlace());
-      auto tmp_eigen =
-          framework::EigenVector<T, Eigen::RowMajor,
-                                 Eigen::DenseIndex>::Flatten(tmp_tensor);
-      auto dim = Eigen::array<int, 1>({{0}});
-      tmp_eigen.device(*place) = (in_x_batch_eigen * outg_batch_eigen).sum(dim);
-      framework::Tensor norm_tmp_tensor;
-      norm_tmp_tensor.mutable_data<T>(framework::make_ddim({1, fea_len}),
-                                      context.GetPlace());
-      auto norm_tmp_eigen =
-          framework::EigenVector<T, Eigen::RowMajor,
-                                 Eigen::DenseIndex>::Flatten(norm_tmp_tensor);
-      norm_tmp_eigen.device(*place) =
-          (x_square_batch_eigen.sum(dim) + epsilon).sqrt();
-      Eigen::array<int, 2> broadcast_dim_col;
-      broadcast_dim_col[1] = 1;
-      broadcast_dim_col[0] = channels;
-      in_g_batch_eigen.device(*place) =
-          in_x_batch_eigen * tmp_eigen.broadcast(broadcast_dim_col);
-      in_g_batch_eigen.device(*place) =
-          in_g_batch_eigen /
-          (norm_tmp_eigen * norm_tmp_eigen).broadcast(broadcast_dim_col);
-      in_g_batch_eigen.device(*place) = outg_batch_eigen - in_g_batch_eigen;
-      // outg_batch_eigen + (in_g_batch_eigen * -1);
-      in_g_batch_eigen.device(*place) =
-          in_g_batch_eigen / norm_tmp_eigen.broadcast(broadcast_dim_col);
-      Eigen::array<int, 2> broadcast_dim_row;
-      broadcast_dim_row[1] = fea_len;
-      broadcast_dim_row[0] = 1;
-      in_g_batch_eigen.device(*place) =
-          in_g_batch_eigen * (scale_eigen.broadcast(broadcast_dim_row));
-    }
+  void Compute(const framework::ExecutionContext& ctx) const override {
+    auto* in_x = ctx.Input<framework::Tensor>("X");
+    auto* in_norm = ctx.Input<framework::Tensor>("Norm");
+    auto* in_dy = ctx.Input<framework::Tensor>(framework::GradVarName("Out"));
+    auto* out_dx = ctx.Output<framework::Tensor>(framework::GradVarName("X"));
+    out_dx->mutable_data<T>(ctx.GetPlace());
+
+    auto xdim = in_x->dims();
+    int axis = ctx.Attr<int>("axis");
+    if (axis < 0) axis = xdim.size() + axis;
+    int pre, n, post;
+    GetDims(xdim, axis, &pre, &n, &post);
+
+    auto* place = ctx.template device_context<DeviceContext>().eigen_device();
+
+    auto x_e = framework::EigenVector<T>::Flatten(*in_x);
+    auto dy_e = framework::EigenVector<T>::Flatten(*in_dy);
+    auto norm_e = framework::EigenVector<T>::Flatten(*in_norm);
+    auto dx_e = framework::EigenVector<T>::Flatten(*out_dx);
+
+    Eigen::DSizes<int, 3> shape(pre, n, post);
+    Eigen::DSizes<int, 2> norm_shape(pre, post);
+    auto x = x_e.reshape(shape);
+    auto dy = dy_e.reshape(shape);
+    auto norm = norm_e.reshape(norm_shape);
+    auto dx = dx_e.reshape(shape);
+
+    framework::Tensor rsum;
+    rsum.mutable_data<T>({pre, post}, ctx.GetPlace());
+    auto sum = framework::EigenTensor<T, 2>::From(rsum);
+
+    Eigen::DSizes<int, 1> rdim(1);
+    Eigen::DSizes<int, 3> bcast(1, n, 1);
+    Eigen::DSizes<int, 3> rshape(pre, 1, post);
+
+    // dx = ( dy/sqrt(sum(x*x)) ) * [1 - x*sum(x) / (sum(x*x) + e)]
+    //    = [dy - dy * x * sum(x) / (sum(x*x) + e)] / sqrt(sum(x*x))
+    //    = [dy - x * sum(x*dy) / (sum(x*x) + e)] / sqrt(sum(x*x))
+    // 1. sum = sum(x*dy)
+    sum.device(*place) = (x * dy).sum(rdim);
+    // 2. dx = x * sum
+    dx.device(*place) = sum.reshape(rshape).broadcast(bcast) * x;
+    // 3. dx / (sum(x*x) + e)
+    // where, norm.pow(2) = sum(x*x) + e, which is calculated in forward.
+    dx.device(*place) = dx / norm.pow(2).broadcast(bcast);
+    // 4. [dy - dx] / sqrt(sum(x*x))
+    dx.device(*place) = (dy - dx) / norm.broadcast(bcast);
   }
 };
 }  // namespace operators

paddle/fluid/operators/slice_op.cc

+/* Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
+
+Licensed 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. */
+
+#include "paddle/fluid/operators/slice_op.h"
+#include <algorithm>
+#include <vector>
+
+namespace paddle {
+namespace operators {
+
+using Tensor = framework::Tensor;
+
+class SliceOp : public framework::OperatorWithKernel {
+ public:
+  using framework::OperatorWithKernel::OperatorWithKernel;
+
+  void InferShape(framework::InferShapeContext *ctx) const override {
+    PADDLE_ENFORCE(ctx->HasInput("Input"),
+                   "Input (Input) of slice op should not be null.");
+    PADDLE_ENFORCE(ctx->HasOutput("Out"),
+                   "Output (Out) of slice op should not be null.");
+
+    auto in_dims = ctx->GetInputDim("Input");
+    PADDLE_ENFORCE(in_dims.size() < 7,
+                   "The rank of input should be less than 7.");
+    framework::DDim out_dims(in_dims);
+    auto axes = ctx->Attrs().Get<std::vector<int>>("axes");
+    auto starts = ctx->Attrs().Get<std::vector<int>>("starts");
+    auto ends = ctx->Attrs().Get<std::vector<int>>("ends");
+
+    PADDLE_ENFORCE_EQ(starts.size(), ends.size());
+    PADDLE_ENFORCE_EQ(starts.size(), axes.size());
+    int dim_value, start, end;
+    for (size_t i = 0; i < axes.size(); ++i) {
+      dim_value = out_dims[axes[i]];
+      start = starts[i] < 0 ? (starts[i] + dim_value) : starts[i];
+      end = ends[i] < 0 ? (ends[i] + dim_value) : ends[i];
+      start = std::max(start, 0);
+      end = std::max(end, 0);
+      start = std::min(start, dim_value);
+      end = std::min(end, dim_value);
+      start = std::min(start, end);
+      out_dims[axes[i]] = end - start;
+    }
+    ctx->SetOutputDim("Out", out_dims);
+  }
+
+ protected:
+  framework::OpKernelType GetExpectedKernelType(
+      const framework::ExecutionContext &ctx) const override {
+    return framework::OpKernelType(
+        framework::ToDataType(ctx.Input<Tensor>("Input")->type()),
+        ctx.GetPlace());
+  }
+};
+
+class SliceOpMaker : public framework::OpProtoAndCheckerMaker {
+ public:
+  void Make() override {
+    AddInput("Input", "Tensor of data to extract slices from.");
+    AddOutput("Out", "Sliced data tensor.");
+
+    AddAttr<std::vector<int>>(
+        "axes",
+        "(list<int>) Axes that `starts` and `ends` apply to. It's optional."
+        "If not present, will be treated as [0, 1, ..., len(`starts`) - 1].");
+    AddAttr<std::vector<int>>(
+        "starts",
+        "(list<int>) Starting indices of corresponding axis in `axes`");
+    AddAttr<std::vector<int>>(
+        "ends",
+        "(list<int>) Starting indices of corresponding axis in `axes`.");
+
+    AddComment(R"DOC(
+Slice Operator.
+
+Produces a slice of the input tensor along multiple axes. Similar to numpy:
+https://docs.scipy.org/doc/numpy/reference/arrays.indexing.html
+Slice uses `axes`, `starts` and `ends` attributes to specify the start and 
+end dimension for each axis in the list of axes, it uses this information
+to slice the input data tensor. If a negative value is passed for any of 
+the start or end indices, it represents number of elements before the end 
+of that dimension. If the value passed to start or end is larger than
+the n (the number of elements in this dimension), it represents n. 
+For slicing to the end of a dimension with unknown size, it is recommended 
+to pass in INT_MAX. If axes are omitted, they are set to [0, ..., ndim-1].
+
+    Example 1:
+    Given:
+        data = [ [1, 2, 3, 4], [5, 6, 7, 8], ]
+        axes = [0, 1]
+        starts = [1, 0]
+        ends = [2, 3]
+    Then:
+        result = [ [5, 6, 7], ]
+
+    Example 2:
+    Given:
+        data = [ [1, 2, 3, 4], [5, 6, 7, 8], ]
+        starts = [0, 1]
+        ends = [-1, 1000]
+    Then:
+        result = [ [2, 3, 4], ]
+)DOC");
+  }
+};
+
+}  // namespace operators
+}  // namespace paddle
+
+namespace ops = paddle::operators;
+REGISTER_OPERATOR(slice, ops::SliceOp, ops::SliceOpMaker,
+                  paddle::framework::EmptyGradOpMaker);
+
+REGISTER_OP_CPU_KERNEL(
+    slice, ops::SliceKernel<paddle::platform::CPUDeviceContext, int>,
+    ops::SliceKernel<paddle::platform::CPUDeviceContext, int64_t>,
+    ops::SliceKernel<paddle::platform::CPUDeviceContext, float>,
+    ops::SliceKernel<paddle::platform::CPUDeviceContext, double>);

paddle/fluid/operators/slice_op.cu

+/* Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
+
+Licensed 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. */
+
+#include "paddle/fluid/operators/slice_op.h"
+
+namespace ops = paddle::operators;
+REGISTER_OP_CUDA_KERNEL(
+    slice, ops::SliceKernel<paddle::platform::CUDADeviceContext, float>,
+    ops::SliceKernel<paddle::platform::CUDADeviceContext, double>,
+    ops::SliceKernel<paddle::platform::CUDADeviceContext, int>,
+    ops::SliceKernel<paddle::platform::CUDADeviceContext, int64_t>);

paddle/fluid/operators/slice_op.h

+/* Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
+
+Licensed 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. */
+
+#pragma once
+#include <algorithm>
+#include <vector>
+#include "paddle/fluid/framework/op_registry.h"
+
+namespace paddle {
+namespace operators {
+
+template <typename DeviceContext, typename T>
+class SliceKernel : public framework::OpKernel<T> {
+ public:
+  void Compute(const framework::ExecutionContext& ctx) const override {
+    int rank = ctx.Input<framework::Tensor>("Input")->dims().size();
+    switch (rank) {
+      case 1:
+        SliceCompute<1>(ctx);
+        break;
+      case 2:
+        SliceCompute<2>(ctx);
+        break;
+      case 3:
+        SliceCompute<3>(ctx);
+        break;
+      case 4:
+        SliceCompute<4>(ctx);
+        break;
+      case 5:
+        SliceCompute<5>(ctx);
+        break;
+      case 6:
+        SliceCompute<6>(ctx);
+        break;
+    }
+  }
+
+ private:
+  template <size_t D>
+  void SliceCompute(const framework::ExecutionContext& context) const {
+    auto& place =
+        *context.template device_context<DeviceContext>().eigen_device();
+    auto in = context.Input<framework::Tensor>("Input");
+    auto out = context.Output<framework::Tensor>("Out");
+    out->mutable_data<T>(context.GetPlace());
+    auto out_dims = out->dims();
+    auto in_dims = in->dims();
+    auto axes = context.Attr<std::vector<int>>("axes");
+    auto starts = context.Attr<std::vector<int>>("starts");
+
+    auto offsets = Eigen::array<int, D>();
+    auto extents = Eigen::array<int, D>();
+    for (size_t i = 0; i < D; ++i) {
+      offsets[i] = 0;
+      extents[i] = out_dims[i];
+    }
+    int start;
+    for (size_t i = 0; i < axes.size(); ++i) {
+      start = starts[i];
+      if (start < 0) {
+        start = (start + in_dims[axes[i]]);
+      }
+      start = std::max(start, 0);
+      offsets[axes[i]] = start;
+    }
+    auto in_t =
+        framework::EigenTensor<T, D, Eigen::RowMajor, Eigen::DenseIndex>::From(
+            *in);
+    auto out_t =
+        framework::EigenTensor<T, D, Eigen::RowMajor, Eigen::DenseIndex>::From(
+            *out);
+    out_t.device(place) = in_t.slice(offsets, extents);
+  }
+};
+}  // namespace operators
+}  // namespace paddle

paddle/fluid/pybind/pybind.cc

@@ -413,6 +413,9 @@ All parameter, weight, gradient are variables in Paddle.
 
   py::class_<framework::Executor>(m, "Executor")
       .def(py::init<const platform::Place &>())
+#ifdef PADDLE_WITH_DISTRIBUTE
+      .def("complete", &Executor::Complete)
+#endif
       .def("run",
            (void (Executor::*)(const ProgramDesc &, Scope *, int, bool, bool)) &
                Executor::Run);

paddle/scripts/paddle_build.sh

@@ -181,6 +181,7 @@ function build() {
     ============================================
 EOF
     make clean
+    make -j `nproc`
     make install -j `nproc`
 }
 

python/paddle/fluid/layers/ops.py

@@ -71,6 +71,7 @@ __all__ = [
     'cumsum',
     'scatter',
     'sum',
+    'slice',
     'polygon_box_transform',
     'shape',
     'maxout',

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

@@ -387,6 +387,12 @@ class TestBook(unittest.TestCase):
             self.assertIsNotNone(output)
         print(str(program))
 
+    def test_l2_normalize(self):
+        program = Program()
+        with program_guard(program):
+            x = layers.data(name='x', shape=[8, 7, 10], dtype="float32")
+            output = layers.l2_normalize(x, axis=1)
+
     def test_maxout(self):
         program = Program()
         with program_guard(program):

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

@@ -17,44 +17,23 @@ import numpy as np
 from op_test import OpTest
 
 
-def norm(input, scale, epsilon):
-    s0, s1, s2, s3 = input.shape
-    x_square = input * input
-    for i in xrange(s0):
-        input_batch = input[i:i + 1, :, :, :]
-        input_batch = input_batch.reshape(s1, s2 * s3)
-        x_square_batch = x_square[i:i + 1, :, :, :]
-        x_square_batch = x_square_batch.reshape(s1, s2 * s3)
-        square_colsum = x_square_batch.sum(axis=0) + epsilon
-        tmp = pow(square_colsum, 0.5)
-        tmp = np.reciprocal(tmp)
-        tmp_tile = np.tile(tmp, s1)
-        tmp_tile = tmp_tile.reshape(s1, s2 * s3)
-        scale_tile = np.tile(scale, (1, s2 * s3))
-        scale_tile = scale_tile.reshape(s1, s2 * s3)
-        out_batch = input_batch * tmp_tile * scale_tile
-        out_batch = out_batch.reshape(1, s1, s2, s3)
-        if i == 0:
-            out = out_batch
-        else:
-            out = np.concatenate((out, out_batch), 0)
-    out.reshape(s0, s1, s2, s3)
-    return out
+def l2_norm(x, axis, epsilon):
+    x2 = x**2
+    s = np.sum(x2, axis=axis, keepdims=True)
+    r = np.sqrt(s + epsilon)
+    y = x / np.broadcast_to(r, x.shape)
+    return y, r
 
 
 class TestNormOp(OpTest):
     def setUp(self):
         self.op_type = "norm"
         self.init_test_case()
-        input = np.random.random(self.shape).astype("float32")
-        scale = np.array([10, 10, 10])
-        self.inputs = {
-            'X': input.astype('float32'),
-            'Scale': scale.astype('float32')
-        }
-        self.attrs = {'epsilon': self.epsilon}
-        output = norm(input, scale, self.epsilon)
-        self.outputs = {'Out': output.astype('float32')}
+        x = np.random.random(self.shape).astype("float64")
+        y, norm = l2_norm(x, self.axis, self.epsilon)
+        self.inputs = {'X': x}
+        self.attrs = {'epsilon': self.epsilon, 'axis': self.axis}
+        self.outputs = {'Out': y, 'Norm': norm}
 
     def test_check_output(self):
         self.check_output()
@@ -63,8 +42,23 @@ class TestNormOp(OpTest):
         self.check_grad(['X'], 'Out')
 
     def init_test_case(self):
-        self.shape = [2, 3, 2, 2]
-        self.epsilon = 1e-6
+        self.shape = [2, 3, 4, 4]
+        self.axis = 1
+        self.epsilon = 1e-8
+
+
+class TestNormOp2(TestNormOp):
+    def init_test_case(self):
+        self.shape = [5, 3, 9, 7]
+        self.axis = 0
+        self.epsilon = 1e-8
+
+
+class TestNormOp3(TestNormOp):
+    def init_test_case(self):
+        self.shape = [5, 3, 2, 7]
+        self.axis = -1
+        self.epsilon = 1e-8
 
 
 if __name__ == '__main__':

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

@@ -70,8 +70,9 @@ class TestNormalization(unittest.TestCase):
     def l2_normalize(self, data, axis, epsilon):
         """ Compute the groundtruth.
         """
-        output = data * np.reciprocal(
-            np.sum(np.square(data), axis=axis, keepdims=True))
+        output = data / np.broadcast_to(
+            np.sqrt(np.sum(np.square(data), axis=axis, keepdims=True)),
+            data.shape)
         return output
 
     def test_l2_normalize(self):

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

+#   Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed 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.
+
+import unittest
+import numpy as np
+from op_test import OpTest
+
+
+class TestSliceOp(OpTest):
+    def setUp(self):
+        self.op_type = "slice"
+        self.config()
+        self.inputs = {'Input': self.input}
+        self.outputs = {'Out': self.out}
+        self.attrs = {
+            'axes': self.axes,
+            'starts': self.starts,
+            'ends': self.ends
+        }
+
+    def config(self):
+        self.input = np.random.random([3, 4, 5, 6]).astype("float32")
+        self.starts = [1, 0, 2]
+        self.ends = [3, 3, 4]
+        self.axes = [0, 1, 2]
+        self.out = self.input[1:3, 0:3, 2:4, :]
+
+    def test_check_output(self):
+        self.check_output()
+
+
+class TestCase1(TestSliceOp):
+    def config(self):
+        self.input = np.random.random([3, 4, 5, 6]).astype("float32")
+        self.starts = [-3, 0, 2]
+        self.ends = [3, 100, -1]
+        self.axes = [0, 1, 2]
+        self.out = self.input[-3:3, 0:100, 2:-1, :]
+
+
+class TestCase2(TestSliceOp):
+    def config(self):
+        self.input = np.random.random([3, 4, 5, 6]).astype("float32")
+        self.starts = [-3, 0, 2]
+        self.ends = [3, 100, -1]
+        self.axes = [0, 1, 3]
+        self.out = self.input[-3:3, 0:100, :, 2:-1]
+
+
+if __name__ == '__main__':
+    unittest.main()

python/paddle/fluid/transpiler/distribute_transpiler.py

@@ -515,35 +515,38 @@ class DistributeTranspiler:
                                        grad_to_block_id, None)
 
         # process distributed lookup_table
-        prefetch_block = None
+        prefetch_var_name_to_block_id = []
         if self.has_distributed_lookup_table:
             pserver_index = self.pserver_endpoints.index(endpoint)
             table_opt_block = self._create_table_optimize_block(
                 pserver_index, pserver_program, pre_block_idx, grad_to_block_id)
-            prefetch_block = self._create_prefetch_block(
+            prefetch_var_name_to_block_id = self._create_prefetch_block(
                 pserver_index, pserver_program, table_opt_block)
 
         # NOTE: if has_distributed_lookup_table is False, then prefetch_block will
         # not be executed, so it's safe to use optimize_block to hold the place
         if self.has_distributed_lookup_table:
-            assert prefetch_block is not None
+            assert len(prefetch_var_name_to_block_id) > 0
         else:
-            assert prefetch_block is None
-            prefetch_block = pserver_program.global_block()
+            assert len(prefetch_var_name_to_block_id) == 0
+
+        attrs = {
+            "OptimizeBlock": pserver_program.block(1),
+            "endpoint": endpoint,
+            "Fanin": self.trainer_num,
+            "sync_mode": self.sync_mode,
+            "grad_to_block_id": grad_to_block_id
+        }
+        if len(prefetch_var_name_to_block_id) > 0:
+            attrs['prefetch_var_name_to_block_id'] \
+                = prefetch_var_name_to_block_id
 
         # step5 append the listen_and_serv op
         pserver_program.global_block().append_op(
             type="listen_and_serv",
             inputs={'X': recv_inputs},
             outputs={},
-            attrs={
-                "OptimizeBlock": pserver_program.block(1),
-                "endpoint": endpoint,
-                "Fanin": self.trainer_num,
-                "PrefetchBlock": prefetch_block,
-                "sync_mode": self.sync_mode,
-                "grad_to_block_id": grad_to_block_id
-            })
+            attrs=attrs)
 
         pserver_program.sync_with_cpp()
         return pserver_program
@@ -608,8 +611,15 @@ class DistributeTranspiler:
     def _replace_lookup_table_op_with_prefetch(self, program,
                                                pserver_endpoints):
         # 1. replace lookup_table_op with split_ids_op -> prefetch_op -> sum_op
-        self.prefetch_input_vars = None
-        self.prefetch_output_vars = None
+        # self.all_prefetch_input_vars =
+        #       [[var0_prefetch_in_pserver0, var0_prefetch_in_pserver1]
+        #        [var1_prefetch_in_pserver0, var1_prefetch_in_pserver1]]
+        self.all_prefetch_input_vars = []
+
+        # self.all_prefetch_input_vars =
+        #       [[var0_prefetch_in_pserver0, var0_prefetch_in_pserver1]
+        #        [var1_prefetch_in_pserver0, var1_prefetch_in_pserver1]]
+        self.all_prefetch_output_vars = []
 
         continue_search_lookup_table_op = True
         while continue_search_lookup_table_op:
@@ -623,18 +633,19 @@ class DistributeTranspiler:
                     ids_name = op.input("Ids")
                     out_name = op.output("Out")
 
-                    if self.prefetch_input_vars is None:
-                        ids_var = program.global_block().vars[ids_name[0]]
-                        self.prefetch_input_vars = self.create_splited_vars(
-                            source_var=ids_var,
-                            block=program.global_block(),
-                            tag="_prefetch_in_")
-                    if self.prefetch_output_vars is None:
-                        out_var = program.global_block().vars[out_name[0]]
-                        self.prefetch_output_vars = self.create_splited_vars(
-                            source_var=out_var,
-                            block=program.global_block(),
-                            tag="_prefetch_out_")
+                    ids_var = program.global_block().vars[ids_name[0]]
+                    prefetch_input_vars = self.create_splited_vars(
+                        source_var=ids_var,
+                        block=program.global_block(),
+                        tag="_prefetch_in_")
+                    self.all_prefetch_input_vars.append(prefetch_input_vars)
+
+                    out_var = program.global_block().vars[out_name[0]]
+                    prefetch_output_vars = self.create_splited_vars(
+                        source_var=out_var,
+                        block=program.global_block(),
+                        tag="_prefetch_out_")
+                    self.all_prefetch_output_vars.append(prefetch_output_vars)
 
                     # insert split_ids_op
                     program.global_block().insert_op(
@@ -646,14 +657,14 @@ class DistributeTranspiler:
                                 for varname in ids_name
                             ]
                         },
-                        outputs={"Out": self.prefetch_input_vars})
+                        outputs={"Out": prefetch_input_vars})
 
                     # insert prefetch_op
                     program.global_block().insert_op(
                         index=lookup_table_op_index + 1,
                         type="prefetch",
-                        inputs={'X': self.prefetch_input_vars},
-                        outputs={"Out": self.prefetch_output_vars},
+                        inputs={'X': prefetch_input_vars},
+                        outputs={"Out": prefetch_output_vars},
                         attrs={
                             "epmap": pserver_endpoints,
                             RPC_OP_ROLE_ATTR_NAME: RPC_OP_ROLE_ATTR_VALUE
@@ -668,7 +679,7 @@ class DistributeTranspiler:
                                 program.global_block().vars[varname]
                                 for varname in ids_name
                             ],
-                            'X': self.prefetch_output_vars
+                            'X': prefetch_output_vars
                         },
                         outputs={
                             "Out": [
@@ -714,30 +725,34 @@ class DistributeTranspiler:
                                optimize_block):
         # STEP: create prefetch block
         table_var = pserver_program.global_block().vars[self.table_name]
-        prefetch_block = pserver_program.create_block(optimize_block.idx)
-        trainer_ids = self.prefetch_input_vars[pserver_index]
-        pserver_ids = pserver_program.global_block().create_var(
-            name=trainer_ids.name,
-            type=trainer_ids.type,
-            shape=trainer_ids.shape,
-            dtype=trainer_ids.dtype)
-        trainer_out = self.prefetch_output_vars[pserver_index]
-        pserver_out = pserver_program.global_block().create_var(
-            name=trainer_out.name,
-            type=trainer_out.type,
-            shape=trainer_out.shape,
-            dtype=trainer_out.dtype)
-        prefetch_block.append_op(
-            type="lookup_sparse_table",
-            inputs={'Ids': pserver_ids,
-                    "W": table_var},
-            outputs={"Out": pserver_out},
-            attrs={
-                "is_sparse": True,  # has no effect on lookup_table op
-                "is_distributed": True,
-                "padding_idx": -1
-            })
-        return prefetch_block
+        prefetch_var_name_to_block_id = []
+        for index in range(len(self.all_prefetch_input_vars)):
+            prefetch_block = pserver_program.create_block(optimize_block.idx)
+            trainer_ids = self.all_prefetch_input_vars[index][pserver_index]
+            pserver_ids = pserver_program.global_block().create_var(
+                name=trainer_ids.name,
+                type=trainer_ids.type,
+                shape=trainer_ids.shape,
+                dtype=trainer_ids.dtype)
+            trainer_out = self.all_prefetch_output_vars[index][pserver_index]
+            pserver_out = pserver_program.global_block().create_var(
+                name=trainer_out.name,
+                type=trainer_out.type,
+                shape=trainer_out.shape,
+                dtype=trainer_out.dtype)
+            prefetch_block.append_op(
+                type="lookup_sparse_table",
+                inputs={'Ids': pserver_ids,
+                        "W": table_var},
+                outputs={"Out": pserver_out},
+                attrs={
+                    "is_sparse": True,  # has no effect on lookup_table op
+                    "is_distributed": True,
+                    "padding_idx": -1
+                })
+            prefetch_var_name_to_block_id.append(trainer_ids.name + ":" + str(
+                prefetch_block.idx))
+        return prefetch_var_name_to_block_id
 
     def _create_table_optimize_block(self, pserver_index, pserver_program,
                                      pre_block_idx, grad_to_block_id):

tools/codestyle/docstring_checker.py

@@ -126,9 +126,10 @@ class DocstringChecker(BaseChecker):
         'W9002':
         ('Doc string does not end with "." period', symbol + "-end-with",
          'Used when a doc string does not end with a period'),
-        'W9003': ('All args with their types must be mentioned in doc string',
-                  symbol + "-with-all-args",
-                  'Used when not all arguments are in the doc string '),
+        'W9003':
+        ('All args with their types must be mentioned in doc string %s',
+         symbol + "-with-all-args",
+         'Used when not all arguments are in the doc string '),
         'W9005': ('Missing docstring or docstring is too short',
                   symbol + "-missing", 'Add docstring longer >=10'),
         'W9006': ('Docstring indent error, use 4 space for indent',
@@ -178,6 +179,8 @@ class DocstringChecker(BaseChecker):
         self.indent_style(node)
 
     def missing_doc_string(self, node):
+        if node.name.startswith("__") or node.name.startswith("_"):
+            return True
         if node.tolineno - node.fromlineno <= 10:
             return True
 
@@ -199,12 +202,16 @@ class DocstringChecker(BaseChecker):
 
         doc = node.doc
         lines = doc.splitlines()
+        line_num = 0
 
         for l in lines:
+            if line_num == 0:
+                continue
             cur_indent = len(l) - len(l.lstrip())
             if cur_indent % indent != 0:
                 self.add_message('W9006', node=node, line=node.fromlineno)
                 return False
+            line_num += 1
 
         return True
 
@@ -320,15 +327,19 @@ class DocstringChecker(BaseChecker):
             return True
 
         parsed_args = doc.args
+        args_not_documented = set(args) - set(parsed_args)
         if len(args) > 0 and len(parsed_args) <= 0:
-            print "debug:parsed args: ", parsed_args
-            self.add_message('W9003', node=node, line=node.fromlineno)
+            self.add_message(
+                'W9003',
+                node=node,
+                line=node.fromlineno,
+                args=list(args_not_documented))
             return False
 
         for t in args:
             if t not in parsed_args:
-                print t, " with (type) not in ", parsed_args
-                self.add_message('W9003', node=node, line=node.fromlineno)
+                self.add_message(
+                    'W9003', node=node, line=node.fromlineno, args=[t, ])
                 return False
 
         return True

tools/codestyle/pylint_pre_commit.hook

@@ -7,13 +7,13 @@ DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )"
 export PYTHONPATH=$DIR:$PYTHONPATH
 
 # The trick to remove deleted files: https://stackoverflow.com/a/2413151
-for file in $(git diff --cached --name-status | awk '$1 != "D" {print $2}'); do
+for file in $(git diff --name-status | awk '$1 != "D" {print $2}'); do
     pylint --disable=all --load-plugins=docstring_checker \
     --enable=doc-string-one-line,doc-string-end-with,doc-string-with-all-args,doc-string-triple-quotes,doc-string-missing,doc-string-indent-error,doc-string-with-returns,doc-string-with-raises $file;
     TOTAL_ERRORS=$(expr $TOTAL_ERRORS + $?);
 done
 
-#exit $TOTAL_ERRORS
+exit $TOTAL_ERRORS
 #For now, just warning:
-exit 0
+#exit 0