Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into add_tensorrt_conv2d_converter

Author: NHZlX

AUTHORS.md

@@ -46,6 +46,7 @@
 | tianbingsz | Tian-Bing Xu |
 | tpatejko | Tomasz Patejko |
 | typhoonzero | Yi Wu |
+| velconia | Qi-Yang Min |
 | wanghaoshuang | Hao-Shuang Wang |
 | wangyang59 | Yang Wang |
 | wangzhen-nlp | Zhen Wang |

doc/fluid/design/ir/draft.md

@@ -1,16 +1,16 @@
 ## Motivation
 
-There is a ```gap``` between the ```Program``` defined by
-user and the ```Executable``` that can be scheduled
+There is a `gap` between the `Program` defined by
+user and the `Executable` that can be scheduled
 efficiently on heterogeneous hardware, either locally
 or distributedly.
 
-Usually, the ```gap``` is bridged by
+Usually, the `gap` is bridged by
 
 * A serious transformations with defined order.
 
 * These transformations usually involve
-```insert, delete, clustering, split, dependency analysis```.
+`insert, delete, clustering, split, dependency analysis`.
 
 * Has a simple way to verify and debug each transformation.
 
@@ -38,44 +38,44 @@ design below.
 
 #### Node
 
-```Node``` represents an operation that performs some computation or
+`Node` represents an operation that performs some computation or
 a variable that is input or output of operation.
 
-```Node```s are connected to other ```Node```s via inputs and outputs.
+`Node`s are connected to other `Node`s via inputs and outputs.
 
 Other properties (maybe device placement information) can be added
-to ```Node``` in the future if it's a
-common requirement of many other ```Pass```es. Otherwise, it should live
-in a ```Node``` wrapper class that is private to some ```Pass``` or be
-a local member of a ```Pass```.
+to `Node` in the future if it's a
+common requirement of many other `Pass`es. Otherwise, it should live
+in a `Node` wrapper class that is private to some `Pass` or be
+a local member of a `Pass`.
 
 #### Graph
 
-```Graph``` contains a list of ```Node```s, which are connected to
+`Graph` contains a list of `Node`s, which are connected to
 each other via inputs and outputs.
 
 TODO: Better definitions for the graph.
 
-```Graph``` can also contain ```Attribute```s. ```Attribute```s
-can be ``any`` thing. For example, it can be a list of "wraper"
-nodes. The ```wrapper``` nodes compose ```Node```s and provide
-helper method for execution or transformation. ```Attribute```
+`Graph` can also contain `Attribute`s. `Attribute`s
+can be `any` thing. For example, it can be a list of "wraper"
+nodes. The `wrapper` nodes compose `Node`s and provide
+helper method for execution or transformation. `Attribute`
 can also contain other things that describe some properties of
-the ```Graph``` or ```Graph``` nodes. ```Attribute``` can be passed
-across ```Pass```. However, it should be used with care.
+the `Graph` or `Graph` nodes. `Attribute` can be passed
+across `Pass`. However, it should be used with care.
 
 #### Pass
 
-```Pass``` represents a transformation of ```Graph```. Its input
-is a ```Graph``` and its output is also a ```Graph```. For example,
-a ```Pass``` can simply print out the ```Graph```. A ```Pass```
-can also fuse some ```Graph```'s ```Node```s.
+`Pass` represents a transformation of `Graph`. Its input
+is a `Graph` and its output is also a `Graph`. For example,
+a `Pass` can simply print out the `Graph`. A `Pass`
+can also fuse some `Graph`'s `Node`s.
 
 #### Optimize
 
-```Optimize``` contains a series of ```Pass``` with defined order.
-```Optimize``` transforms a ```Graph``` that only contains raw
-modeling logic to a ```Graph``` that can be run efficiently while
+`Optimize` contains a series of `Pass` with defined order.
+`Optimize` transforms a `Graph` that only contains raw
+modeling logic to a `Graph` that can be run efficiently while
 maintaining the original modeling logic.
 
 

paddle/fluid/API.spec

@@ -35,8 +35,7 @@ paddle.fluid.program_guard ArgSpec(args=[], varargs='args', keywords='kwds', def
 paddle.fluid.get_var ArgSpec(args=['name', 'program'], varargs=None, keywords=None, defaults=(None,))
 paddle.fluid.Executor.__init__ ArgSpec(args=['self', 'place'], varargs=None, keywords=None, defaults=None)
 paddle.fluid.Executor.as_lodtensor ArgSpec(args=['self', 'data'], varargs=None, keywords=None, defaults=None)
-paddle.fluid.Executor.begin_pass ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None)
-paddle.fluid.Executor.end_pass ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None)
+paddle.fluid.Executor.close ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None)
 paddle.fluid.Executor.run ArgSpec(args=['self', 'program', 'feed', 'fetch_list', 'feed_var_name', 'fetch_var_name', 'scope', 'return_numpy', 'use_program_cache'], varargs=None, keywords=None, defaults=(None, None, None, 'feed', 'fetch', None, True, False))
 paddle.fluid.global_scope ArgSpec(args=[], varargs=None, keywords=None, defaults=None)
 paddle.fluid.scope_guard ArgSpec(args=[], varargs='args', keywords='kwds', defaults=None)
@@ -200,31 +199,23 @@ paddle.fluid.layers.argsort ArgSpec(args=['input', 'axis', 'name'], varargs=None
 paddle.fluid.layers.ones ArgSpec(args=['shape', 'dtype', 'force_cpu'], varargs=None, keywords=None, defaults=(False,))
 paddle.fluid.layers.zeros ArgSpec(args=['shape', 'dtype', 'force_cpu'], varargs=None, keywords=None, defaults=(False,))
 paddle.fluid.layers.reverse ArgSpec(args=['x', 'axis'], varargs=None, keywords=None, defaults=None)
-paddle.fluid.layers.split_lod_tensor ArgSpec(args=['input', 'mask', 'level'], varargs=None, keywords=None, defaults=(0,))
-paddle.fluid.layers.merge_lod_tensor ArgSpec(args=['in_true', 'in_false', 'x', 'mask', 'level'], varargs=None, keywords=None, defaults=(0,))
 paddle.fluid.layers.While.__init__ ArgSpec(args=['self', 'cond', 'name'], varargs=None, keywords=None, defaults=(None,))
 paddle.fluid.layers.While.block ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None)
 paddle.fluid.layers.While.complete ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None)
 paddle.fluid.layers.Switch.__init__ ArgSpec(args=['self', 'name'], varargs=None, keywords=None, defaults=(None,))
 paddle.fluid.layers.Switch.case ArgSpec(args=['self', 'condition'], varargs=None, keywords=None, defaults=None)
 paddle.fluid.layers.Switch.default ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None)
-paddle.fluid.layers.lod_rank_table ArgSpec(args=['x', 'level'], varargs=None, keywords=None, defaults=(0,))
-paddle.fluid.layers.max_sequence_len ArgSpec(args=['rank_table'], varargs=None, keywords=None, defaults=None)
-paddle.fluid.layers.lod_tensor_to_array ArgSpec(args=['x', 'table'], varargs=None, keywords=None, defaults=None)
-paddle.fluid.layers.array_to_lod_tensor ArgSpec(args=['x', 'table'], varargs=None, keywords=None, defaults=None)
 paddle.fluid.layers.increment ArgSpec(args=['x', 'value', 'in_place'], varargs=None, keywords=None, defaults=(1.0, True))
 paddle.fluid.layers.array_write ArgSpec(args=['x', 'i', 'array'], varargs=None, keywords=None, defaults=(None,))
 paddle.fluid.layers.create_array ArgSpec(args=['dtype'], varargs=None, keywords=None, defaults=None)
 paddle.fluid.layers.less_than ArgSpec(args=['x', 'y', 'force_cpu', 'cond'], varargs=None, keywords='ignored', defaults=(None, None))
 paddle.fluid.layers.equal ArgSpec(args=['x', 'y', 'cond'], varargs=None, keywords='ignored', defaults=(None,))
 paddle.fluid.layers.array_read ArgSpec(args=['array', 'i'], varargs=None, keywords=None, defaults=None)
-paddle.fluid.layers.shrink_memory ArgSpec(args=['x', 'i', 'table'], varargs=None, keywords=None, defaults=None)
 paddle.fluid.layers.array_length ArgSpec(args=['array'], varargs=None, keywords=None, defaults=None)
 paddle.fluid.layers.IfElse.__init__ ArgSpec(args=['self', 'cond', 'name'], varargs=None, keywords=None, defaults=(None,))
 paddle.fluid.layers.IfElse.false_block ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None)
 paddle.fluid.layers.IfElse.input ArgSpec(args=['self', 'x'], varargs=None, keywords=None, defaults=None)
 paddle.fluid.layers.IfElse.output ArgSpec(args=['self'], varargs='outs', keywords=None, defaults=None)
-paddle.fluid.layers.IfElse.parent_block ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None)
 paddle.fluid.layers.IfElse.true_block ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None)
 paddle.fluid.layers.DynamicRNN.__init__ ArgSpec(args=['self', 'name'], varargs=None, keywords=None, defaults=(None,))
 paddle.fluid.layers.DynamicRNN.block ArgSpec(args=[], varargs='args', keywords='kwds', defaults=None)
@@ -233,9 +224,6 @@ paddle.fluid.layers.DynamicRNN.output ArgSpec(args=['self'], varargs='outputs',
 paddle.fluid.layers.DynamicRNN.static_input ArgSpec(args=['self', 'x'], varargs=None, keywords=None, defaults=None)
 paddle.fluid.layers.DynamicRNN.step_input ArgSpec(args=['self', 'x'], varargs=None, keywords=None, defaults=None)
 paddle.fluid.layers.DynamicRNN.update_memory ArgSpec(args=['self', 'ex_mem', 'new_mem'], varargs=None, keywords=None, defaults=None)
-paddle.fluid.layers.ConditionalBlock.__init__ ArgSpec(args=['self', 'inputs', 'is_scalar_condition', 'name'], varargs=None, keywords=None, defaults=(False, None))
-paddle.fluid.layers.ConditionalBlock.block ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None)
-paddle.fluid.layers.ConditionalBlock.complete ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None)
 paddle.fluid.layers.StaticRNN.__init__ ArgSpec(args=['self', 'name'], varargs=None, keywords=None, defaults=(None,))
 paddle.fluid.layers.StaticRNN.complete_op ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None)
 paddle.fluid.layers.StaticRNN.memory ArgSpec(args=['self', 'init', 'shape', 'batch_ref', 'init_value', 'init_batch_dim_idx', 'ref_batch_dim_idx'], varargs=None, keywords=None, defaults=(None, None, None, 0.0, 0, 1))

paddle/fluid/framework/CMakeLists.txt

@@ -22,7 +22,12 @@ endif()
 
 cc_test(eigen_test SRCS eigen_test.cc DEPS tensor)
 
-nv_test(mixed_vector_test SRCS mixed_vector_test.cu DEPS place memory device_context tensor)
+if(WITH_GPU)
+  nv_test(mixed_vector_test SRCS mixed_vector_test.cc mixed_vector_test.cu DEPS place memory device_context tensor)
+else()
+  cc_test(mixed_vector_test SRCS mixed_vector_test.cc DEPS place memory device_context tensor)
+endif()
+
 cc_library(lod_tensor SRCS lod_tensor.cc DEPS ddim place tensor framework_proto recordio)
 cc_test(lod_tensor_test SRCS lod_tensor_test.cc DEPS lod_tensor memory)
 nv_test(lod_tensor_gpu_test SRCS lod_tensor_test.cu DEPS lod_tensor)

paddle/fluid/framework/block_desc.h

@@ -88,9 +88,8 @@ class BlockDesc {
   OpDesc *InsertOp(size_t index);
 
   /*
-   * Remove Op and its input/output variables.
-   * Note that for either input or output variable, if it is also an input or
-   * output variable of other ops, we should remain it.
+   * Only remove op itself,
+   * do nothing to its input and output variables
    */
   void RemoveOp(size_t s, size_t e);
 

paddle/fluid/framework/details/CMakeLists.txt

@@ -1,11 +1,11 @@
-cc_library(var_handle SRCS var_handle.cc DEPS place framework_proto)
+cc_library(var_handle SRCS var_handle.cc DEPS place framework_proto node)
 cc_library(op_handle_base SRCS op_handle_base.cc DEPS var_handle device_context lod_tensor)
 cc_library(scale_loss_grad_op_handle SRCS scale_loss_grad_op_handle.cc DEPS op_handle_base scope lod_tensor ddim memory)
 cc_library(fetch_op_handle SRCS fetch_op_handle.cc DEPS op_handle_base scope lod_tensor ddim memory)
 cc_library(computation_op_handle SRCS computation_op_handle.cc DEPS framework_proto scope place operator op_registry)
 cc_library(rpc_op_handle SRCS rpc_op_handle.cc DEPS framework_proto scope place operator op_registry)
 
-cc_library(ssa_graph_builder SRCS ssa_graph_builder.cc DEPS graph)
+cc_library(ssa_graph_builder SRCS ssa_graph_builder.cc DEPS graph graph_helper)
 cc_library(ssa_graph_printer SRCS ssa_graph_printer.cc DEPS ssa_graph_builder)
 cc_library(ssa_graph_checker SRCS ssa_graph_checker.cc DEPS ssa_graph_builder)