diff options
Diffstat (limited to 'nerv')
78 files changed, 4065 insertions, 2055 deletions
diff --git a/nerv/Makefile b/nerv/Makefile index c0db53a..68465a1 100644 --- a/nerv/Makefile +++ b/nerv/Makefile @@ -1,3 +1,11 @@ +ifndef LUA_BINDIR +$(error Please build the package via luarocks: `luarocks make`) +endif + +ifndef CUDA_BASE +$(error CUDA_BASE is not set) +endif + .PHONY: build install clean SHELL := /bin/bash @@ -6,14 +14,15 @@ LIB_PATH := $(LUA_BINDIR)/../lib INC_PATH := $(LUA_BINDIR)/../include/nerv LUA_DIR = $(INST_LUADIR)/nerv OBJ_DIR := $(BUILD_DIR)/objs -ISUBDIR := io matrix luaT +ISUBDIR := lib matrix lib/io lib/matrix lib/luaT SUBDIR := matrix io layer examples nn tnn lib/io lib/luaT lib/matrix INC_SUBDIR := $(addprefix $(INC_PATH)/,$(ISUBDIR)) OBJ_SUBDIR := $(addprefix $(OBJ_DIR)/,$(SUBDIR)) LUA_SUBDIR := $(addprefix $(LUA_DIR)/,$(SUBDIR)) -INCS := common.h matrix/matrix.h io/chunk_file.h luaT/luaT.h +INCS := lib/common.h lib/matrix/matrix.h lib/matrix/mmatrix.h lib/io/chunk_file.h lib/luaT/luaT.h \ + matrix/matrix.h CORE_OBJS := lib/common.o lib/io/chunk_file.o \ lib/matrix/mmatrix.o lib/matrix/cumatrix.o lib/matrix/cukernel.o NERV_OBJS := nerv.o \ @@ -33,17 +42,17 @@ LUA_LIBS := matrix/init.lua io/init.lua init.lua \ layer/init.lua layer/affine.lua layer/sigmoid.lua layer/tanh.lua layer/softmax_ce.lua layer/softmax.lua \ layer/window.lua layer/bias.lua layer/combiner.lua layer/mse.lua \ layer/elem_mul.lua layer/lstm.lua layer/lstm_gate.lua layer/dropout.lua layer/gru.lua \ - nn/init.lua nn/layer_repo.lua nn/param_repo.lua nn/layer_dag.lua \ - io/sgd_buffer.lua \ - tnn/init.lua tnn/sutil.lua tnn/tnn.lua + layer/graph.lua layer/rnn.lua layer/duplicate.lua layer/identity.lua \ + nn/init.lua nn/layer_repo.lua nn/param_repo.lua nn/network.lua \ + io/sgd_buffer.lua io/seq_buffer.lua INCLUDE := -I $(LUA_INCDIR) -DLUA_USE_APICHECK -#CUDA_BASE := /usr/local/cuda-7.0 -CUDA_BASE := /usr/local/cuda CUDA_INCLUDE := -I $(CUDA_BASE)/include/ INCLUDE += $(CUDA_INCLUDE) -LDFLAGS := -L$(CUDA_BASE)/lib64/ -Wl,-rpath=$(CUDA_BASE)/lib64/ -lcudart -lcublas -lcurand +CUDA_LDFLAGS := -L$(CUDA_BASE)/lib64/ -Wl,-rpath=$(CUDA_BASE)/lib64/ -lcudart -lcuda -lcublas -lcurand +override CFLAGS += $(NERV_FEAT) + NVCC := $(CUDA_BASE)/bin/nvcc EMPTY := SPACE := $(EMPTY) $(EMPTY) @@ -64,11 +73,11 @@ $(LUA_DIR)/%.lua: %.lua cp $< $@ $(LIB_PATH)/libnervcore.so: $(CORE_OBJS) - gcc -shared -o $@ $^ $(LDFLAGS) -lcblas + gcc -shared -o $@ $^ $(LDFLAGS) $(CUDA_LDFLAGS) $(BLAS_LDFLAGS) $(LIB_PATH)/libluaT.so: $(LUAT_OBJS) - gcc -shared -o $@ $^ $(LDFLAGS) + gcc -shared -o $@ $^ $(INST_LIBDIR)/libnerv.so: $(NERV_OBJS) $(LIB_PATH)/libnervcore.so $(LIB_PATH)/libluaT.so - gcc -shared -o $@ $(NERV_OBJS) $(LDFLAGS) -Wl,-rpath=$(LIB_PATH) -L$(LIB_PATH) -lnervcore -lluaT + gcc -shared -o $@ $(NERV_OBJS) -Wl,-rpath=$(LIB_PATH) -L$(LIB_PATH) -lnervcore -lluaT $(OBJ_DIR)/matrix/cumatrix.o: matrix/generic/cumatrix.c matrix/generic/matrix.c $(OBJ_DIR)/matrix/mmatrix.o: matrix/generic/mmatrix.c matrix/generic/matrix.c @@ -82,5 +91,5 @@ clean: install: $(LIBS) $(LUA_DIR) $(LUA_SUBDIR) $(LUA_LIBS) $(INC_SUBDIR) $(INCS) -$(INC_PATH)/%.h: lib/%.h +$(INC_PATH)/%.h: %.h cp $< $@ diff --git a/nerv/doc/nerv.md b/nerv/doc/nerv.md index 28411f5..125928d 100644 --- a/nerv/doc/nerv.md +++ b/nerv/doc/nerv.md @@ -1,6 +1,6 @@ -#The Nerv utility functions# +# The Nerv utility functions Part of the [Nerv](../README.md) toolkit. -##Methods## +## Methods * __string = nerv.typename(obj a)__ A registered function, the original function is `luaT_lua_typename`. In some cases if you call `type(a)` for object of some class in __Nerv__(like __Nerv.CuMatrix__) it will only return "userdata"(because it is created in C), in this case you can use this method to get its type. @@ -14,4 +14,4 @@ A registered function, the original function is `luaT_newmetatable`, it returns * __string = nerv.setmetatable(table self, string tname)__ A registered function, the original function is `luaT_lua_setmetatable`. It assigns the metatable registered in __luaT__ by the name *tname* to the table *self*. And return *tname* to user. * __table = nerv.get_type(string typename)__ -Returns the type(`loadstring("return " .. typename)`).
\ No newline at end of file +Returns the type(`loadstring("return " .. typename)`). diff --git a/nerv/doc/nerv_class.md b/nerv/doc/nerv_class.md index 99f63e7..8314b12 100644 --- a/nerv/doc/nerv_class.md +++ b/nerv/doc/nerv_class.md @@ -1,10 +1,10 @@ -#The Nerv OOP# +# The Nerv OOP Part of the [Nerv](../README.md) toolkit. -##Methods## +## Methods * __metatable mt, metatable mpt = nerv.class(string tname, string parenttname)__ This method is used to create a class by the name `tname`, which inherits `parenttname` in __Nerv__, then you create a new instance of this class by calling `obj=tname(...)`. The `tname.__init(...)` method(if defined) will be called in the constructing. The metatable of the class and its parent class will be returned. -##Examples## +## Examples * This example implements a simple `nerv.Counter` class which is inherited by `nerv.BetterCounter`. ``` @@ -33,4 +33,4 @@ c1 = nerv.Counter(1) print(c1.c) bc1 = nerv.BetterCounter(1, 1) print(bc1.c, bc1.bc) -```
\ No newline at end of file +``` diff --git a/nerv/doc/nerv_io.md b/nerv/doc/nerv_io.md index 07589df..299362f 100644 --- a/nerv/doc/nerv_io.md +++ b/nerv/doc/nerv_io.md @@ -1,7 +1,7 @@ -#The Nerv IO Package# +# The Nerv IO Package Part of the [Nerv](../README.md) toolkit. -##Description## +## Description The main class that the user uses to store and read parameter object to and from files is __nerv.ChunkFile__. In the file, a parameter object will be saved using a standard format. First is the length(in byte) of this object, then a table which includes some meta information of the object, and a data area. Below is an example text file. ``` @@ -23,7 +23,7 @@ In the file, a parameter object will be saved using a standard format. First is 3.000000 3.000000 3.000000 ``` -##Methods## +## Methods * __ChunkFile ChunkFile(string fn, string mode)__ `mode` can be `r` or `w`, for reading or writing a file. The returned __ChunkFile__ will be ready to write or read objects which follows the __nerv.Param__ interface(using `write_chunk` and `read_chunk`). * __void ChunkFile.write_chunk(ChunkFile self, Param p)__ @@ -33,7 +33,7 @@ Read the __Param__ object by id `id` from the file `self`. It will be constructe * __void ChunkFile.close(ChunkFile self)__ Close the opened file. -##Examples## +## Examples * An example showing how to use __ChunkFile__ to store and read parameter objects. ``` require 'io' @@ -96,7 +96,7 @@ do end ``` -##Developer Notes## +## Developer Notes * There are four classes in to deal with chunk data, which are __nerv.ChunkFile__, __nerv.ChunkFileHandle__, __nerv.ChunkInfo__, __nerv.ChunkData__. Below is the underlying C structs. ``` typedef struct ChunkFileHandle { @@ -110,4 +110,5 @@ typedef struct ChunkData { char *data; } ChunkData; ``` -* In __Nerv.io__, a returned(by `ChunkFile.__init`) __nerv.ChunkFile__ will have a member `handle`, which is a __nerv.ChunkFileHandle__.
\ No newline at end of file + +* In __Nerv.io__, a returned(by `ChunkFile.__init`) __nerv.ChunkFile__ will have a member `handle`, which is a __nerv.ChunkFileHandle__. diff --git a/nerv/doc/nerv_layer.md b/nerv/doc/nerv_layer.md index de2fb12..dd7c9bb 100644 --- a/nerv/doc/nerv_layer.md +++ b/nerv/doc/nerv_layer.md @@ -1,9 +1,9 @@ -#The Nerv Layer Package# +# The Nerv Layer Package Part of the [Nerv](../README.md) toolkit. -##Description## +## Description __nerv.Layer__ is the base class and most of its methods are abstract. -###Class hierarchy and their members### +### Class hierarchy and their members * __nerv.Layer__. * `table dim_in` It specifies the dimensions of the inputs. * `table dim_out` It specifies the dimensions of the outputs. @@ -20,7 +20,7 @@ __nerv.Layer__ is the base class and most of its methods are abstract. * `int total_frams` Records how many frames have passed. * `bool compressed` The reference distribution can be a one-hot format. This feature is enabled by `layer_conf.compressed`. -##Methods## +## Methods * __void Layer.\_\_init(Layer self, string id, table global_conf, table layer_conf)__ Abstract method. The constructing method should assign `id` to `self.id` and `global_conf` to `self.gconf`, `layer_conf.dim_in` to `self.dim_in`, `layer_conf.dim_out` to `self.dim_out`. `dim_in` and `dim_out` are a list specifies the dimensions of the inputs and outputs. Also, `layer_conf` will include the parameters, which should also be properly saved. @@ -43,7 +43,7 @@ Check whether `#self.dim_in == len_in` and `#self.dim_out == len_out`, if violat Abstract method. The layer should return a list containing its parameters. -####nerv.Layer.get\_dim(self)#### +#### nerv.Layer.get\_dim(self) * Returns: `dim_in`: __table__. `dim_out`: __table__. @@ -52,7 +52,7 @@ The layer should return a list containing its parameters. * Description: Returns `self.dim_in, self.dim_out`. -##Examples## +## Examples * a basic example using __Nerv__ layers to a linear classification. ``` @@ -178,3 +178,4 @@ for l = 0, 10, 1 do end --[[end training]]-- ``` + diff --git a/nerv/doc/nerv_matrix.md b/nerv/doc/nerv_matrix.md index dfd843d..3782eb3 100644 --- a/nerv/doc/nerv_matrix.md +++ b/nerv/doc/nerv_matrix.md @@ -1,8 +1,8 @@ -#The Nerv Matrix Package# +# The Nerv Matrix Package Part of the [Nerv](../README.md) toolkit. -##Description## -###Underlying structure### +## Description +### Underlying structure In the begining is could be useful to know something about the underlying structure of a __Nerv__ matrix. Please keep in mind that matrice in __Nerv__ is row-major. Every matrix object is a encapsulation of a C struct that describes the attributes of this matrix. ``` @@ -20,12 +20,12 @@ typedef struct Matrix { It is worth mentioning that that `data_ref` is a counter which counts the number of references to its memory space, mind that it will also be increased when a row of the matrix is referenced(`col = m[2]`). A __Nerv__ matrix will deallocate its space when this counter is decreased to zero. Also note that all assigning operation in __Nerv__ is reference copy, you can use `copy_tod` or `copy_toh` method to copy value. Also, row assigning operations like `m1[2]=m2[3]` is forbidden in __Nerv__. -###Class hierarchy### +### Class hierarchy The class hierarchy of the matrix classes can be clearly observed in `matrix/init.c`. First there is a abstract base class __Nerv.Matrix__, which is inherited by __Nerv.CuMatrix__ and __Nerv.MMatrix__(also abstract). Finally, there is __Nerv.CuMatrixFloat__, __Nerv.CuMatrixDouble__, inheriting __Nerv.CuMatrix__, and __Nerv.MMatrixFloat__, __Nerv.MMatrixDouble__, __Nerv.MMatrixInt__ , inheriting __Nerv.MMatrix__. -##Methods## +## Methods Mind that usually a matrix object can only do calculation with matrix of its own type(a __Nerv.CuMatrixFloat__ matrix can only do add operation with a __Nerv.CuMatrixFloat__). In the methods description below, __Matrix__ could be __Nerv.CuMatrixFloat__, __Nerv.CuMatrixDouble__, __Nerv.MMatrixFloat__ or __Nerv.MMatrixDouble__. __Element_type__ could be `float` or `double`, respectively. * __Matrix = Matrix(int nrow, int ncol)__ @@ -53,6 +53,8 @@ Return a new __Matrix__ of size (1,`self.ncol`), which stores the sum of all col Return a new __Matrix__ of size (`self.nrow`,1), which stores the sum of all rows of __Matrix__ `self`. * __Matrix Matrix.rowmax(Matrix self)__ Return a new __Matrix__ of size (`self.nrow`,1), which stores the max value of all rows of __Matrix__ `self`. +* __Matrix Matrix.rowmax_idx(Matrix self)__ +Return two new __Matrix__ of size (`self.nrow`,1), which stores the max value of all rows of __Matrix__ `self`, and its corresponding column indices(start from zero). * __Matrix Matrix.trans(Matrix self)__ Return a new __Matrix__ of size (`self.ncol`,`self.nrow`), which stores the transpose of __Matrix__ `self`. * __void Matrix.copy_fromh(Matrix self, MMatrix a)__ @@ -81,8 +83,8 @@ Fill the content of __Matrix__ `self` to be `value`. Set the element of __Matrix__ `self` to be elementwise-sigmoid of `ma`. * __void Matrix.sigmoid_grad(Matrix self, Matrix err, Matrix output)__ Set the element of __Matrix__ `self`, to be `self[i][j]=err[i][j]*output[i][j]*(1-output[i][j])`. This function is used to propagate sigmoid layer error. -* __void Matrix.softmax(Matrix self, Matrix a)__ -Calculate a row-by-row softmax of __Matrix__ `a` and save the result in `self`. +* __Matrix Matrix.softmax(Matrix self, Matrix a)__ +Calculate a row-by-row softmax of __Matrix__ `a` and save the result in `self`. Returns a new `self.nrow*1` index matrix that stores the index of the maximum value of each row. * __void Matrix.mul_elem(Matrix self, Matrix ma, Matrix mb)__ Calculate element-wise multiplication of __Matrix__ `ma` and `mb`, store the result in `self`. * __void Matrix.log_elem(Matrix self, Matrix ma)__ @@ -113,7 +115,7 @@ Write `self` to the file position in `chunk`. * __void MMatrix.copy_from(MMatrix ma, MMatrix mb,[int b_bgein, int b_end, int a_begin])__ Copy a part of `mb`(rows of index `[b_begin..b_end)`) to `ma` beginning at row index `a_begin`. If not specified, `b_begin` will be `0`, `b_end` will be `b.nrow`, `a_begin` will be `0`. -##Examples## +## Examples * Use `get_dataref_value` to test __Nerv__'s matrix space allocation. ``` m = 10 @@ -134,6 +136,7 @@ print("test fm:get_dataref_value:", fm:get_dataref_value()) print(fm) print(dm) ``` + * Test some __Matrix__ calculations. ``` m = 4 @@ -167,3 +170,4 @@ print(a) a:log_elem(fs) print(a) ``` + diff --git a/nerv/doc/nerv_nn.md b/nerv/doc/nerv_nn.md index c57447d..63537fb 100644 --- a/nerv/doc/nerv_nn.md +++ b/nerv/doc/nerv_nn.m |