Merge remote-tracking branch 'upstream/master'

Conflicts:
	.gitignore
	Makefile

.gitignore

@@ -83,13 +83,15 @@ R-package/inst/*
 *.bin
 
 # ipython notebook
-example/notebooks/.ipynb_checkpoints/*
 *_pb2.py
+*.ipynb_checkpoints*
+input.txt*
 
 # Jetbrain
 .idea
 
 # ctags
+<<<<<<< HEAD
 tags
 
 

CONTRIBUTORS.md

@@ -92,3 +92,7 @@ List of Contributors
 * [Junru Shao](https://github.com/yzgysjr)
 * [Xiao Liu](https://github.com/skylook)
 * [Lowik CHANUSSOT](https://github.com/Nzeuwik)
+* [Alexander Skidanov](https://github.com/SkidanovAlex)
+* [Ruixiang Zhang](https://github.com/sodabeta7)
+* [Lodewic van Twillert](https://github.com/Lodewic)
+* [Aditya Kumar](https://github.com/hiraditya)

Makefile

@@ -92,14 +92,16 @@ ifeq ($(USE_DIST_KVSTORE), 1)
 	LDFLAGS += $(PS_LDFLAGS_A)
 endif
 
+include $(MXNET_PLUGINS)
+
 .PHONY: clean all test lint doc clean_all rcpplint rcppexport roxygen
 
 all: lib/libmxnet.a lib/libmxnet.so $(BIN)
 
 SRC = $(wildcard src/*.cc src/*/*.cc)
-OBJ = $(patsubst src/%.cc, build/%.o, $(SRC))
+OBJ = $(patsubst %.cc, build/%.o, $(SRC))
 CUSRC = $(wildcard src/*/*.cu)
-CUOBJ = $(patsubst src/%.cu, build/%_gpu.o, $(CUSRC))
+CUOBJ = $(patsubst %.cu, build/%_gpu.o, $(CUSRC))
 
 ifneq ($(EXTRA_OPERATORS),)
 	EXTRA_SRC = $(wildcard $(EXTRA_OPERATORS)/*.cc $(EXTRA_OPERATORS)/*/*.cc)
@@ -113,10 +115,27 @@ else
 	EXTRA_CUOBJ =
 endif
 
+# plugin
+ifeq ($(USE_TORCH), 1)
+	CFLAGS += -I$(TORCH_PATH)/install/include -I$(TORCH_PATH)/install/include/TH -I$(TORCH_PATH)/install/include/THC -DMXNET_USE_TORCH=1
+	LDFLAGS += -L$(TORCH_PATH)/install/lib -lluajit -lluaT -lTH -lTHC -L$(TORCH_PATH)/install/lib/lua/5.1 -lpaths -ltorch -lnn
+	ifeq ($(USE_CUDA), 1)
+		LDFLAGS += -lcutorch -lcunn
+	endif
+
+	TORCH_SRC = $(wildcard plugin/torch/*.cc)
+	PLUGIN_OBJ += $(patsubst %.cc, build/%.o, $(TORCH_SRC))
+	TORCH_CUSRC = $(wildcard plugin/torch/*.cu)
+	PLUGIN_CUOBJ += $(patsubst %.cu, build/%_gpu.o, $(TORCH_CUSRC))
+else
+	CFLAGS += -DMXNET_USE_TORCH=0
+endif
+
 LIB_DEP += $(DMLC_CORE)/libdmlc.a
-ALL_DEP = $(OBJ) $(EXTRA_OBJ) $(LIB_DEP) $(config)
+ALL_DEP = $(OBJ) $(EXTRA_OBJ) $(PLUGIN_OBJ) $(LIB_DEP)
 ifeq ($(USE_CUDA), 1)
-	ALL_DEP += $(CUOBJ) $(EXTRA_CUOBJ)
+	ALL_DEP += $(CUOBJ) $(EXTRA_CUOBJ) $(PLUGIN_CUOBJ)
+	LDFLAGS += -lcuda
 endif
 
 ifeq ($(USE_NVRTC), 1)
@@ -127,14 +146,25 @@ else
 endif
 
 
-build/%.o: src/%.cc
+build/src/%.o: src/%.cc
+	@mkdir -p $(@D)
+	$(CXX) -std=c++0x $(CFLAGS) -MM -MT build/src/$*.o $< >build/src/$*.d
+	$(CXX) -std=c++0x -c $(CFLAGS) -c $< -o $@
+
+build/src/%_gpu.o: src/%.cu
+	@mkdir -p $(@D)
+	$(NVCC) $(NVCCFLAGS) -Xcompiler "$(CFLAGS)" -M -MT build/src/$*_gpu.o $< >build/src/$*_gpu.d
+	$(NVCC) -c -o $@ $(NVCCFLAGS) -Xcompiler "$(CFLAGS)" $<
+
+build/plugin/%.o: plugin/%.cc
 	@mkdir -p $(@D)
-	$(CXX) -std=c++0x $(CFLAGS) -MM -MT build/$*.o $< >build/$*.d
+	$(CXX) -std=c++0x $(CFLAGS) -MM -MT build/plugin/$*.o $< >build/plugin/$*.d
 	$(CXX) -std=c++0x -c $(CFLAGS) -c $< -o $@
 
-build/%_gpu.o: src/%.cu
+# A nvcc bug cause this to generate "generic/xxx.h" dependencies from torch headers.
+# $(NVCC) $(NVCCFLAGS) -Xcompiler "$(CFLAGS)" -M -MT build/plugin/$*_gpu.o $< >build/plugin/$*_gpu.d
+build/plugin/%_gpu.o: plugin/%.cu
 	@mkdir -p $(@D)
-	$(NVCC) $(NVCCFLAGS) -Xcompiler "$(CFLAGS)" -M -MT build/$*_gpu.o $< >build/$*_gpu.d
 	$(NVCC) -c -o $@ $(NVCCFLAGS) -Xcompiler "$(CFLAGS)" $<
 
 $(EXTRA_OPERATORS)/build/%.o: $(EXTRA_OPERATORS)/%.cc
@@ -147,6 +177,8 @@ $(EXTRA_OPERATORS)/build/%_gpu.o: $(EXTRA_OPERATORS)/%.cu
 	$(NVCC) $(NVCCFLAGS) -Xcompiler "$(CFLAGS) -Isrc/operator" -M -MT $(EXTRA_OPERATORS)/build/$*_gpu.o $< >$(EXTRA_OPERATORS)/build/$*_gpu.d
 	$(NVCC) -c -o $@ $(NVCCFLAGS) -Xcompiler "$(CFLAGS) -Isrc/operator" $<
 
+# NOTE: to statically link libmxnet.a we need the option
+# --Wl,--whole-archive -lmxnet --Wl,--no-whole-archive
 lib/libmxnet.a: $(ALL_DEP)
 	@mkdir -p $(@D)
 	ar crv $@ $(filter %.o, $?)
@@ -161,7 +193,6 @@ lib/libmxnet.so: $(ALL_DEP) copy_cuda_deps
 	@mkdir -p $(@D)
 	$(CXX) $(CFLAGS) -shared -o $@ $(filter %.o %.a, $^) $(LDFLAGS)
 
-# ps-lite
 $(PS_PATH)/build/libps.a:
 	$(MAKE) CXX=$(CXX) DEPS_PATH=$(DEPS_PATH) -C $(PS_PATH) ps
 	ln -fs $(PS_PATH)/tracker .
@@ -180,7 +211,7 @@ include tests/cpp/unittest.mk
 test: $(TEST)
 
 lint: rcpplint
-	python2 dmlc-core/scripts/lint.py mxnet ${LINT_LANG} include src scripts python predict/python
+	python2 dmlc-core/scripts/lint.py mxnet ${LINT_LANG} include src plugin scripts python predict/python
 
 doc: doxygen
 
@@ -206,15 +237,24 @@ rpkg:	roxygen
 	cp -rf dmlc-core/include/* R-package/inst/include/
 	R CMD build --no-build-vignettes R-package
 
+ifneq ($(EXTRA_OPERATORS),)
 clean:
 	$(RM) -r build lib bin *~ */*~ */*/*~ */*/*/*~
 	cd $(DMLC_CORE); make clean; cd -
 	cd $(PS_PATH); make clean; cd -
+	$(RM) -r $(EXTRA_OPERATORS)/build
+else
+clean:
+	$(RM) -r build lib bin *~ */*~ */*/*~ */*/*/*~
+	cd $(DMLC_CORE); make clean; cd -
+	cd $(PS_PATH); make clean; cd -
+endif
 
 clean_all: clean
 
 -include build/*.d
 -include build/*/*.d
+-include build/*/*/*.d
 ifneq ($(EXTRA_OPERATORS),)
 	-include $(EXTRA_OPERATORS)/build/*.d
 endif

R-package/DESCRIPTION

@@ -14,7 +14,7 @@ BugReports: https://github.com/dmlc/mxnet/issues
 Imports:
     methods,
     Rcpp (>= 0.12.1),
-    DiagrammeR,
+    DiagrammeR (>= 0.8.1),
     data.table,
     jsonlite,
     magrittr,

R-package/NAMESPACE

@@ -44,6 +44,7 @@ export(mx.model.FeedForward.create)
 export(mx.model.load)
 export(mx.model.save)
 export(mx.nd.abs)
+export(mx.nd.argmax.channel)
 export(mx.nd.array)
 export(mx.nd.ceil)
 export(mx.nd.choose.element.0index)
@@ -77,8 +78,10 @@ export(mx.simple.bind)
 export(mx.symbol.Activation)
 export(mx.symbol.BatchNorm)
 export(mx.symbol.BlockGrad)
+export(mx.symbol.Cast)
 export(mx.symbol.Concat)
 export(mx.symbol.Convolution)
+export(mx.symbol.Crop)
 export(mx.symbol.Deconvolution)
 export(mx.symbol.Dropout)
 export(mx.symbol.ElementWiseSum)

R-package/R/mxnet_generated.R

@@ -12,6 +12,16 @@
 #' @name mx.nd.abs
 NULL
 
+#' Take sum of the src.The result will be ndarray of shape (1,) on the same device.
+#' 
+#' @param src  NDArray
+#'     Source input to the function
+#' @return out The result mx.ndarray
+#' 
+#' @export
+#' @name mx.nd.argmax.channel
+NULL
+
 #' Take ceil value of the src
 #' 
 #' @param src  NDArray
@@ -269,12 +279,20 @@ mx.io.CSVIter <- function(...) {
 #'     Augmentation Param: Maxmum image size after resizing.
 #' @param min.img.size  float, optional, default=0
 #'     Augmentation Param: Minimum image size after resizing.
+#' @param random.h  int, optional, default='0'
+#'     Augmentation Param: Maximum value of H channel in HSL color space.
+#' @param random.s  int, optional, default='0'
+#'     Augmentation Param: Maximum value of S channel in HSL color space.
+#' @param random.l  int, optional, default='0'
+#'     Augmentation Param: Maximum value of L channel in HSL color space.
 #' @param rotate  int, optional, default='-1'
 #'     Augmentation Param: Rotate angle.
 #' @param fill.value  int, optional, default='255'
 #'     Augmentation Param: Maximum value of illumination variation.
 #' @param data.shape  Shape(tuple), required
 #'     Dataset Param: Shape of each instance generated by the DataIter.
+#' @param inter.method  int, optional, default='1'
+#'     Augmentation Param: 0-NN 1-bilinear 2-cubic 3-area 4-lanczos4 9-auto 10-rand.
 #' @param mirror  boolean, optional, default=False
 #'     Augmentation Param: Whether to mirror the image.
 #' @param rand.mirror  boolean, optional, default=False
@@ -378,6 +396,21 @@ mx.symbol.BlockGrad <- function(...) {
   mx.varg.symbol.BlockGrad(list(...))
 }
 
+#' Cast array to a different data type.
+#' 
+#' @param data  Symbol
+#'     Input data to cast function.
+#' @param dtype  {'float16', 'float32', 'float64', 'int32', 'uint8'}, required
+#'     Target data type.
+#' @param name  string, optional
+#'     Name of the resulting symbol.
+#' @return out The result mx.symbol
+#' 
+#' @export
+mx.symbol.Cast <- function(...) {
+  mx.varg.symbol.Cast(list(...))
+}
+
 #' Perform an feature concat on channel dim (dim 1) over all the inputs.
 #' 
 #' @param num.args  int, required
@@ -405,6 +438,8 @@ mx.symbol.Concat <- function(...) {
 #'     convolution kernel size: (y, x)
 #' @param stride  Shape(tuple), optional, default=(1, 1)
 #'     convolution stride: (y, x)
+#' @param dilate  Shape(tuple), optional, default=(1, 1)
+#'     convolution dilate: (y, x)
 #' @param pad  Shape(tuple), optional, default=(0, 0)
 #'     pad for convolution: (y, x)
 #' @param num.filter  int (non-negative), required
@@ -424,6 +459,25 @@ mx.symbol.Convolution <- function(...) {
   mx.varg.symbol.Convolution(list(...))
 }
 
+#' Crop the 2th and 3th dim of input data, with the corresponding size of w_h orwith widht and height of the second input symbol
+#' 
+#' @param num.args  int, required
+#'     Number of inputs for crop, if equals one, then we will use the h_wfor crop heihgt and width, else if equals two, then we will use the heightand width of the second input symbol, we name crop_like here
+#' @param offset  Shape(tuple), optional, default=(0, 0)
+#'     corp offset coordinate: (y, x)
+#' @param h.w  Shape(tuple), optional, default=(0, 0)
+#'     corp height and weight: (h, w)
+#' @param center.crop  boolean, optional, default=False
+#'     If set to true, then it will use be the center_crop,or it will crop using the shape of crop_like
+#' @param name  string, optional
+#'     Name of the resulting symbol.
+#' @return out The result mx.symbol
+#' 
+#' @export
+mx.symbol.Crop <- function(...) {
+  mx.varg.symbol.Crop(list(...))
+}
+
 #' Apply deconvolution to input then add a bias.
 #' 
 #' @param data  Symbol
@@ -704,8 +758,12 @@ mx.symbol.SliceChannel <- function(...) {
 #'     Input data to softmax.
 #' @param grad.scale  float, optional, default=1
 #'     Scale the gradient by a float factor
+#' @param ignore.label  float, optional, default=-1
+#'     the ignore_label will not work in backward, and this onlybe used when multi_output=true
 #' @param multi.output  boolean, optional, default=False
 #'     If set to true, for a (n,k,x_1,..,x_n) dimensionalinput tensor, softmax will generate n*x_1*...*x_n output, eachhas k classes
+#' @param use.ignore  boolean, optional, default=False
+#'     If set to true, the ignore_label value will not contributorto the backward gradient
 #' @param name  string, optional
 #'     Name of the resulting symbol.
 #' @return out The result mx.symbol
@@ -736,8 +794,12 @@ mx.symbol.SoftmaxActivation <- function(...) {
 #'     Input data to softmax.
 #' @param grad.scale  float, optional, default=1
 #'     Scale the gradient by a float factor
+#' @param ignore.label  float, optional, default=-1
+#'     the ignore_label will not work in backward, and this onlybe used when multi_output=true
 #' @param multi.output  boolean, optional, default=False
 #'     If set to true, for a (n,k,x_1,..,x_n) dimensionalinput tensor, softmax will generate n*x_1*...*x_n output, eachhas k classes
+#' @param use.ignore  boolean, optional, default=False
+#'     If set to true, the ignore_label value will not contributorto the backward gradient
 #' @param name  string, optional
 #'     Name of the resulting symbol.
 #' @return out The result mx.symbol
@@ -772,6 +834,8 @@ mx.symbol.SwapAxis <- function(...) {
 #'     Input filter. Only used by nearest sample_type.
 #' @param sample.type  {'bilinear', 'nearest'}, required
 #'     upsampling method
+#' @param multi.input.mode  {'concat', 'sum'},optional, default='concat'
+#'     How to handle multiple input. concat means concatenate upsampled images along the channel dimension. sum means add all images together, only available for nearest neighbor upsampling.
 #' @param num.args  int, required
 #'     Number of inputs to be upsampled. For nearest neighbor upsampling, this can be 1-N; the size of output will be(scale*h_0,scale*w_0) and all other inputs will be upsampled to thesame size. For bilinear upsampling this must be 2; 1 input and 1 weight.
 #' @param name  string, optional