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

… argmin_argmax
PaddlePaddle · Jun 14, 2018 · fa002cd · fa002cd
2 parents e0f883e + 9169b3b
commit fa002cd
Show file tree

Hide file tree

Showing 94 changed files with 3,134 additions and 934 deletions.
diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -61,6 +61,7 @@ option(EIGEN_USE_THREADS "Compile with multi-threaded Eigen"            OFF)
 option(WITH_ARM_FP16    "Use half precision support on armv8.2-a cpu"   OFF)
 option(WITH_FAST_BUNDLE_TEST    "Bundle tests that can be run in a single process together to reduce launch overhead"   OFF)
 option(WITH_CONTRIB     "Compile the third-party contributation"        OFF)
+option(WITH_ANAKIN      "Compile with Anakin library"                   OFF)
 option(WITH_GRPC     "Use grpc as the default rpc framework"            ${WITH_DISTRIBUTE})
 
 # CMAKE_BUILD_TYPE
@@ -193,7 +194,10 @@ set(EXTERNAL_LIBS
 if(WITH_GPU)
     include(cuda)
     include(tensorrt)
-endif(WITH_GPU)
+    include(external/anakin)
+else()
+  set(WITH_ANAKIN OFF CACHE STRING "Anakin is valid only when GPU is set." FORCE)
+endif()
 
 if(WITH_AMD_GPU)
     find_package(HIP)

diff --git a/benchmark/fluid/fluid_benchmark.py b/benchmark/fluid/fluid_benchmark.py
@@ -188,7 +188,7 @@ def train(avg_loss, infer_prog, optimizer, train_reader, test_reader, batch_acc,
         print_train_time(start_time, time.time(), num_samples)
         print("Pass: %d, Loss: %f" % (pass_id, np.mean(train_losses))),
         # evaluation
-        if not args.no_test and batch_acc:
+        if not args.no_test and batch_acc and not args.use_reader_op:
             pass_test_acc = test(exe, infer_prog, test_reader, feeder,
                                  batch_acc)
             print(", Test Accuracy: %f" % pass_test_acc)
@@ -285,11 +285,12 @@ def train_parallel(avg_loss, infer_prog, optimizer, train_reader, test_reader,
             batch_id += 1
 
         print_train_time(start_time, time.time(), num_samples)
-        if not args.no_test and batch_acc:
+        if not args.no_test and batch_acc and not args.use_reader_op:
+            # we have not implement record io for test
+            # skip test when use args.use_reader_op
             test_acc = test(startup_exe, infer_prog, test_reader, feeder,
                             batch_acc)
             print("Pass: %d, Test Accuracy: %f\n" % (pass_id, test_acc))
-        exit(0)
 
 
 def print_arguments(args):

diff --git a/benchmark/fluid/models/resnet.py b/benchmark/fluid/models/resnet.py
@@ -199,7 +199,10 @@ def get_model(args):
     batched_train_reader = paddle.batch(
         paddle.reader.shuffle(
             train_reader, buf_size=5120),
-        batch_size=args.batch_size * args.gpus)
-    batched_test_reader = paddle.batch(train_reader, batch_size=args.batch_size)
+        batch_size=args.batch_size * args.gpus,
+        drop_last=True)
+    batched_test_reader = paddle.batch(
+        train_reader, batch_size=args.batch_size, drop_last=True)
 
-    return avg_cost, inference_program, optimizer, batched_train_reader, batched_test_reader, batch_acc
+    return avg_cost, inference_program, optimizer, batched_train_reader,\
+                   batched_test_reader, batch_acc
diff --git a/cmake/configure.cmake b/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

diff --git a/cmake/external/anakin.cmake b/cmake/external/anakin.cmake
@@ -0,0 +1,42 @@
+if (NOT WITH_ANAKIN)
+  return()
+endif()
+
+set(ANAKIN_INSTALL_DIR "${THIRD_PARTY_PATH}/install/anakin" CACHE PATH
+  "Anakin install path." FORCE)
+set(ANAKIN_INCLUDE "${ANAKIN_INSTALL_DIR}" CACHE STRING "root of Anakin header files")
+set(ANAKIN_LIBRARY "${ANAKIN_INSTALL_DIR}" CACHE STRING "path of Anakin library")
+
+set(ANAKIN_COMPILE_EXTRA_FLAGS -Wno-error=unused-variable -Wno-error=format-extra-args -Wno-error=comment -Wno-error=format -Wno-error=switch -Wno-error=return-type -Wno-error=non-virtual-dtor -Wno-reorder -Wno-error=cpp)
+
+set(ANAKIN_LIBRARY_URL "https://github.com/pangge/Anakin/releases/download/3.0/anakin_release_simple.tar.gz")
+
+# A helper function used in Anakin, currently, to use it, one need to recursively include
+# nearly all the header files.
+function(fetch_include_recursively root_dir)
+    if (IS_DIRECTORY ${root_dir})
+        include_directories(${root_dir})
+    endif()
+
+    file(GLOB ALL_SUB RELATIVE ${root_dir} ${root_dir}/*)
+    foreach(sub ${ALL_SUB})
+        if (IS_DIRECTORY ${root_dir}/${sub})
+            fetch_include_recursively(${root_dir}/${sub})
+        endif()
+    endforeach()
+endfunction()
+
+# download library
+message(STATUS "Download Anakin library from ${ANAKIN_LIBRARY_URL}")
+execute_process(COMMAND bash -c "mkdir -p ${ANAKIN_INSTALL_DIR}")
+execute_process(COMMAND bash -c "rm -rf ${ANAKIN_INSTALL_DIR}/*")
+execute_process(COMMAND bash -c "cd ${ANAKIN_INSTALL_DIR}; wget -q ${ANAKIN_LIBRARY_URL}")
+execute_process(COMMAND bash -c "mkdir -p ${ANAKIN_INSTALL_DIR}")
+execute_process(COMMAND bash -c "cd ${ANAKIN_INSTALL_DIR}; tar xzf anakin_release_simple.tar.gz")
+
+if (WITH_ANAKIN)
+    message(STATUS "Anakin for inference is enabled")
+    message(STATUS "Anakin is set INCLUDE:${ANAKIN_INCLUDE} LIBRARY:${ANAKIN_LIBRARY}")
+    fetch_include_recursively(${ANAKIN_INCLUDE})
+    link_directories(${ANAKIN_LIBRARY})
+endif()
diff --git a/cmake/external/openblas.cmake b/cmake/external/openblas.cmake
@@ -29,6 +29,8 @@ IF(NOT ${CBLAS_FOUND})
         "${CBLAS_INSTALL_DIR}/lib/${CMAKE_STATIC_LIBRARY_PREFIX}openblas${CMAKE_STATIC_LIBRARY_SUFFIX}"
         CACHE FILEPATH "openblas library." FORCE)
 
+    ADD_DEFINITIONS(-DPADDLE_USE_OPENBLAS)
+
     SET(OPENBLAS_CC "${CMAKE_C_COMPILER} -Wno-unused-but-set-variable -Wno-unused-variable")
     SET(OPENBLAS_COMMIT "v0.2.20")
 

diff --git a/doc/fluid/api/detection.rst b/doc/fluid/api/detection.rst
diff --git a/doc/fluid/api/gen_doc.sh b/doc/fluid/api/gen_doc.sh
@@ -1,5 +1,5 @@
 #!/bin/bash
-python gen_doc.py layers --submodules control_flow device io nn ops tensor > layers.rst
+python gen_doc.py layers --submodules control_flow device io nn ops tensor detection learning_rate_scheduler > layers.rst
 
 for module in data_feeder clip metrics executor initializer io nets optimizer param_attr profiler regularizer
 do

diff --git a/doc/fluid/api/io.rst b/doc/fluid/api/io.rst
@@ -59,21 +59,3 @@ get_inference_program
 ..  autofunction:: paddle.fluid.io.get_inference_program
     :noindex:
 
-save_checkpoint
----------------
-
-..  autofunction:: paddle.fluid.io.save_checkpoint
-    :noindex:
-
-load_checkpoint
----------------
-
-..  autofunction:: paddle.fluid.io.load_checkpoint
-    :noindex:
-
-clean_checkpoint
-----------------
-
-..  autofunction:: paddle.fluid.io.clean_checkpoint
-    :noindex:
-
diff --git a/doc/fluid/api/layers.rst b/doc/fluid/api/layers.rst
@@ -181,12 +181,6 @@ Print
 ..  autofunction:: paddle.fluid.layers.Print
     :noindex:
 
-is_empty
---------
-
-..  autofunction:: paddle.fluid.layers.is_empty
-    :noindex:
-
 device
 ======
 
@@ -261,19 +255,6 @@ double_buffer
 ..  autofunction:: paddle.fluid.layers.double_buffer
     :noindex:
 
-random_data_generator
----------------------
-
-..  autofunction:: paddle.fluid.layers.random_data_generator
-    :noindex:
-
-Preprocessor
-------------
-
-..  autoclass:: paddle.fluid.layers.Preprocessor
-    :members:
-    :noindex:
-
 nn
 ==
 
@@ -613,30 +594,6 @@ roi_pool
 ..  autofunction:: paddle.fluid.layers.roi_pool
     :noindex:
 
-dice_loss
----------
-
-..  autofunction:: paddle.fluid.layers.dice_loss
-    :noindex:
-
-resize_bilinear
----------------
-
-..  autofunction:: paddle.fluid.layers.resize_bilinear
-    :noindex:
-
-gather
-------
-
-..  autofunction:: paddle.fluid.layers.gather
-    :noindex:
-
-random_crop
------------
-
-..  autofunction:: paddle.fluid.layers.random_crop
-    :noindex:
-
 ops
 ===
 
@@ -784,12 +741,6 @@ sum
 ..  autofunction:: paddle.fluid.layers.sum
     :noindex:
 
-shape
------
-
-..  autofunction:: paddle.fluid.layers.shape
-    :noindex:
-
 sigmoid
 -------
 
@@ -1039,3 +990,93 @@ zeros
 ..  autofunction:: paddle.fluid.layers.zeros
     :noindex:
 
+detection
+=========
+
+multi_box_head
+--------------
+
+..  autofunction:: paddle.fluid.layers.multi_box_head
+    :noindex:
+
+bipartite_match
+---------------
+
+..  autofunction:: paddle.fluid.layers.bipartite_match
+    :noindex:
+
+target_assign
+-------------
+
+..  autofunction:: paddle.fluid.layers.target_assign
+    :noindex:
+
+detection_output
+----------------
+
+..  autofunction:: paddle.fluid.layers.detection_output
+    :noindex:
+
+ssd_loss
+--------
+
+..  autofunction:: paddle.fluid.layers.ssd_loss
+    :noindex:
+
+detection_map
+-------------
+
+..  autofunction:: paddle.fluid.layers.detection_map
+    :noindex:
+
+iou_similarity
+--------------
+
+..  autofunction:: paddle.fluid.layers.iou_similarity
+    :noindex:
+
+box_coder
+---------
+
+..  autofunction:: paddle.fluid.layers.box_coder
+    :noindex:
+
+learning_rate_scheduler
+=======================
+
+exponential_decay
+-----------------
+
+..  autofunction:: paddle.fluid.layers.exponential_decay
+    :noindex:
+
+natural_exp_decay
+-----------------
+
+..  autofunction:: paddle.fluid.layers.natural_exp_decay
+    :noindex:
+
+inverse_time_decay
+------------------
+
+..  autofunction:: paddle.fluid.layers.inverse_time_decay
+    :noindex:
+
+polynomial_decay
+----------------
+
+..  autofunction:: paddle.fluid.layers.polynomial_decay
+    :noindex:
+
+piecewise_decay
+---------------
+
+..  autofunction:: paddle.fluid.layers.piecewise_decay
+    :noindex:
+
+noam_decay
+----------
+
+..  autofunction:: paddle.fluid.layers.noam_decay
+    :noindex:
+
diff --git a/doc/fluid/api/optimizer.rst b/doc/fluid/api/optimizer.rst
@@ -89,13 +89,6 @@ DecayedAdagradOptimizer
     :members:
     :noindex:
 
-RMSPropOptimizer
-----------------
-
-..  autoclass:: paddle.fluid.optimizer.RMSPropOptimizer
-    :members:
-    :noindex:
-
 Adadelta
 --------
 

diff --git a/doc/fluid/api/profiler.rst b/doc/fluid/api/profiler.rst
@@ -23,15 +23,3 @@ profiler
 ..  autofunction:: paddle.fluid.profiler.profiler
     :noindex:
 
-start_profiler
---------------
-
-..  autofunction:: paddle.fluid.profiler.start_profiler
-    :noindex:
-
-stop_profiler
--------------
-
-..  autofunction:: paddle.fluid.profiler.stop_profiler
-    :noindex:
-
diff --git a/doc/survey/dynamic_graph.md b/doc/survey/dynamic_graph.md
@@ -171,7 +171,7 @@ Pytorch chooses immediate evaluation. It avoids ever materializing a "forward gr
 
 ## What can fluid learn from them?
 
-TBD
+Please refer to `paddle/contrib/dynamic/`.
 
 # Appendix
 

diff --git a/doc/v2/api/config/evaluators.rst b/doc/v2/api/config/evaluators.rst
@@ -101,7 +101,7 @@ value_printer
     :noindex:
 
 Detection
-=====
+==========
 
 detection_map
 -------------
-Original file line number
+Diff line change
@@ Expand Up @@
     ## What can fluid learn from them?
-    TBD
+    Please refer to `paddle/contrib/dynamic/`.
     # Appendix
@@ Expand Down @@