...

6 files changed, 63 insertions, 14 deletions

diff --git a/Makefile b/Makefile
index 086527a..531e43f 100644
--- a/Makefile
+++ b/Makefile
@@ -1,13 +1,14 @@
 .PHONY: tnet
+INCLUDE := -I $(BUILD_DIR)/luajit-2.0/include/luajit-2.0
 all: tnet $(OBJ_DIR)/tnet_io/test $(LIB_DIR)/libspeech.so
 $(LIB_DIR)/libspeech.so: $(OBJ_DIR)/tnet_io/cwrapper.o $(OBJ_DIR)/init.o $(OBJ_DIR)/tnet_io/init.o $(OBJ_DIR)/tnet_io/libKaldiLib.a
 	gcc -shared -o $@ $(OBJ_DIR)/tnet_io/cwrapper.o $(OBJ_DIR)/init.o $(OBJ_DIR)/tnet_io/libKaldiLib.a $(OBJ_DIR)/tnet_io/init.o -lstdc++ -Wl,-rpath=$(LIB_DIR) -L$(LIB_DIR) -lnerv
 $(OBJ_DIR)/tnet_io/test: $(OBJ_DIR)/tnet_io/cwrapper.o $(OBJ_DIR)/tnet_io/test.o $(OBJ_DIR)/tnet_io/libKaldiLib.a
 	gcc -o $@ $^ -Wl,-rpath=$(LIB_DIR) -L$(LIB_DIR) -lnerv -Wl,-rpath=$(BUILD_DIR)/luajit-2.0/lib -L$(BUILD_DIR)/luajit-2.0/lib -lluajit-5.1 -lstdc++ -lm
 $(OBJ_DIR)/tnet_io/cwrapper.o: tnet_io/cwrapper.cpp
-	g++ -o $@ -c $< -DHAVE_ATLAS -I tnet_io/KaldiLib/ -g -fPIC
+	g++ -o $@ -c $< -DHAVE_ATLAS -I tnet_io/KaldiLib/ -g -fPIC $(INCLUDE)
 $(OBJ_DIR)/%.o: %.c
-	gcc -o $@ -c $< -g -I $(BUILD_DIR)/luajit-2.0/include/luajit-2.0 -fPIC
+	gcc -o $@ -c $< -g $(INCLUDE) -fPIC
 clean:
 	-rm $(OBJ_DIR)/tnet_io/*.o
 	$(MAKE) -C tnet_io/KaldiLib/ clean
diff --git a/examples/tnet_preprocessing_example.lua b/examples/tnet_preprocessing_example.lua
new file mode 100644
index 0000000..c36463b
--- /dev/null
+++ b/examples/tnet_preprocessing_example.lua
@@ -0,0 +1,46 @@
+require 'libspeech'
+require 'layer.affine'
+frm_ext = 5
+feat_repo = nerv.TNetFeatureRepo(
+                                "/slfs1/users/mfy43/swb_ivec/train_bp.scp",
+                                "/slfs1/users/mfy43/swb_ivec/plp_0_d_a.conf",
+                                frm_ext)
+lab_repo = nerv.TNetLabelRepo(
+                                "/slfs1/users/mfy43/swb_ivec/ref.mlf",
+                                "map",
+                                "/slfs1/users/mfy43/swb_ivec/dict",
+                                "*/",
+                                "lab")
+feat_utter = feat_repo:cur_utter()
+-- print(feat_utter)
+-- lab_utter = lab_repo:get_utter(feat_repo, feat_utter:nrow() - frm_ext * 2)
+-- print(lab_utter)
+gconf = {mat_type = nerv.CuMatrixFloat}
+cf = nerv.ChunkFile("global_transf.nerv", "r")
+bias1 = cf:read_chunk("bias1", gconf)
+window1 = cf:read_chunk("window1", gconf)
+bias2 = cf:read_chunk("bias2", gconf)
+window2 = cf:read_chunk("window2", gconf)
+
+cf2 = nerv.ChunkFile("input.param", "r")
+input = cf2:read_chunk("input", gconf)
+
+step = frm_ext * 2 + 1
+expanded = nerv.CuMatrixFloat(feat_utter:nrow(), feat_utter:ncol() * step)
+expanded:expand_frm(nerv.CuMatrixFloat.new_from_host(feat_utter), frm_ext)
+
+rearranged = expanded:create()
+rearranged:rearrange_frm(expanded, step)
+
+rearranged:add_row(bias1.trans, 1)
+rearranged:scale_row(window1.trans)
+rearranged:add_row(bias2.trans, 1)
+rearranged:scale_row(window2.trans)
+
+for i = 0, 9 do
+    row_diff = input.trans[i] - rearranged[i + 5]
+    for j = 0, row_diff:ncol() - 1 do
+        nerv.utils.printf("%.8f ", row_diff[j])
+    end
+    nerv.utils.printf("\n")
+end
diff --git a/tnet_io/cwrapper.cpp b/tnet_io/cwrapper.cpp
index df4bdef..e82f3f8 100644
--- a/tnet_io/cwrapper.cpp
+++ b/tnet_io/cwrapper.cpp
@@ -2,6 +2,7 @@
 #include "KaldiLib/Labels.h"
 #include "KaldiLib/Common.h"
 #include "KaldiLib/UserInterface.h"
+#include "../../common.h"
 #include <string>
 #define SNAME "TNET"
 
@@ -9,7 +10,7 @@ extern "C" {
 #include "cwrapper.h"
 #include "string.h"
 
-    extern Matrix *nerv_matrix_host_float_new_(long nrow, long ncol);
+    extern Matrix *nerv_matrix_host_float_new_(lua_State *L, long nrow, long ncol);
 
     struct TNetFeatureRepo {
         TNet::FeatureRepository feature_repo;
@@ -33,7 +34,6 @@ extern "C" {
     TNetFeatureRepo *tnet_feature_repo_new(const char *p_script, const char *config, int context) {
         TNetFeatureRepo *repo = new TNetFeatureRepo();
         repo->ui.ReadConfig(config);
-        fprintf(stderr, "read config\n");
         repo->swap_features = !repo->ui.GetBool(SNAME":NATURALREADORDER", TNet::IsBigEndian());
         /* load defaults */
         repo->target_kind = repo->ui.GetFeatureParams(&repo->deriv_order,
@@ -53,14 +53,14 @@ extern "C" {
         return repo;
     }
 
-    Matrix *tnet_feature_repo_read_utterance(TNetFeatureRepo *repo) {
+    Matrix *tnet_feature_repo_read_utterance(TNetFeatureRepo *repo, lua_State *L) {
         Matrix *mat;                    /* nerv implementation */
         repo->feature_repo.ReadFullMatrix(repo->feats_host);
         std::string utter_str = repo->feature_repo.Current().Logical();
         repo->feats_host.CheckData(utter_str);
         int n = repo->feats_host.Rows();
         int m = repo->feats_host.Cols();
-        mat = nerv_matrix_host_float_new_(n, m);
+        mat = nerv_matrix_host_float_new_(L, n, m);
         size_t stride = mat->stride;
         fprintf(stderr, "[tnet] feature: %s %d %d\n", utter_str.c_str(), n, m);
         for (int i = 0; i < n; i++)
@@ -68,7 +68,7 @@ extern "C" {
             float *row = repo->feats_host.pRowData(i);
             float *nerv_row = (float *)((char *)mat->data.f + i * stride);
             /* use memmove to copy the row, since KaldiLib uses compact storage */
-             memmove(nerv_row, row, sizeof(float) * m);
+            memmove(nerv_row, row, sizeof(float) * m);
         }
         return mat;
     }
@@ -110,14 +110,15 @@ extern "C" {
     Matrix *tnet_label_repo_read_utterance(TNetLabelRepo *repo,
                                             size_t frames,
                                             size_t sample_rate,
-                                            const char *tag) {
+                                            const char *tag,
+                                            lua_State *L) {
         std::vector<TNet::Matrix<float> >  labs_hosts; /* KaldiLib implementation */
         Matrix *mat;
         repo->label_repo.GenDesiredMatrixExt(labs_hosts, frames,
                                             sample_rate, tag);
         int n = labs_hosts[0].Rows();
         int m = labs_hosts[0].Cols();
-        mat = nerv_matrix_host_float_new_(n, m);
+        mat = nerv_matrix_host_float_new_(L, n, m);
         size_t stride = mat->stride;
         fprintf(stderr, "[tnet] label: %s %d %d\n", tag, n, m);
         for (int i = 0; i < n; i++)
diff --git a/tnet_io/cwrapper.h b/tnet_io/cwrapper.h
index 7a9449a..810901b 100644
--- a/tnet_io/cwrapper.h
+++ b/tnet_io/cwrapper.h
@@ -9,7 +9,7 @@ extern "C" {
 
     TNetFeatureRepo *tnet_feature_repo_new(const char *scp,
                                             const char *config, int context);
-    Matrix *tnet_feature_repo_read_utterance(TNetFeatureRepo *repo);
+    Matrix *tnet_feature_repo_read_utterance(TNetFeatureRepo *repo, lua_State *L);
     size_t tnet_feature_repo_current_samplerate(TNetFeatureRepo *repo);
     const char *tnet_feature_repo_current_tag(TNetFeatureRepo *repo);
     void tnet_feature_repo_next(TNetFeatureRepo *repo);
@@ -24,7 +24,8 @@ extern "C" {
     Matrix *tnet_label_repo_read_utterance(TNetLabelRepo *repo,
                                             size_t frames,
                                             size_t sample_rate,
-                                            const char *tag);
+                                            const char *tag,
+                                            lua_State *L);
 
     void tnet_label_repo_destroy(TNetLabelRepo *repo);
 #ifdef __cplusplus
diff --git a/tnet_io/init.c b/tnet_io/init.c
index 7195eb7..3fa7cb8 100644
--- a/tnet_io/init.c
+++ b/tnet_io/init.c
@@ -29,7 +29,7 @@ static int feat_repo_current_tag(lua_State *L) {
 
 static int feat_repo_current_utterance(lua_State *L) {
     TNetFeatureRepo *repo = luaT_checkudata(L, 1, nerv_tnet_feat_repo_tname);
-    Matrix *utter = tnet_feature_repo_read_utterance(repo);
+    Matrix *utter = tnet_feature_repo_read_utterance(repo, L);
     luaT_pushudata(L, utter, nerv_matrix_host_float_tname);
     return 1;
 }
@@ -67,7 +67,7 @@ static int label_repo_read_utterance(lua_State *L) {
     Matrix *utter = tnet_label_repo_read_utterance(repo,
                         frames,
                         tnet_feature_repo_current_samplerate(feat_repo),
-                        tnet_feature_repo_current_tag(feat_repo));
+                        tnet_feature_repo_current_tag(feat_repo), L);
     luaT_pushudata(L, utter, nerv_matrix_host_float_tname);
     return 1;
 }
diff --git a/tnet_io/tnet.mk b/tnet_io/tnet.mk
index ba4862b..9f933db 100644
--- a/tnet_io/tnet.mk
+++ b/tnet_io/tnet.mk
@@ -40,7 +40,7 @@ AS = $(CROSS_COMPILE)as
 ##############################################################
 ##### PATH TO CUDA TOOLKIT #####
 #CUDA_TK_BASE=/usr/local/share/cuda-3.2.12
-CUDA_TK_BASE=/usr/local/cuda
+#CUDA_TK_BASE=/usr/local/cuda
 ##### PATH TO CUDA TOOLKIT #####