diff options
| -rw-r--r-- | htk_io/init.lua | 3 | ||||
| -rw-r--r-- | htk_io/tools/nerv_to_tnet.lua | 60 | ||||
| -rw-r--r-- | htk_io/tools/tnet_to_nerv.c | 2 | ||||
| -rw-r--r-- | htk_io/tools/tnet_to_nerv.cpp | 8 | ||||
| -rw-r--r-- | kaldi_io/example/swb_baseline_basic.lua | 157 | ||||
| -rw-r--r-- | kaldi_io/kaldi.mk | 70 |
6 files changed, 297 insertions, 3 deletions
diff --git a/htk_io/init.lua b/htk_io/init.lua index af92140..c4dfff9 100644 --- a/htk_io/init.lua +++ b/htk_io/init.lua @@ -34,7 +34,8 @@ function TNetReader:get_data() -- global transf if self.global_transf ~= nil then feat_utter = nerv.speech_utils.global_transf(feat_utter, - self.global_transf, self.frm_ext, self.gconf) + self.global_transf, self.frm_ext, + self.frm_ext, self.gconf) end res[self.feat_id] = feat_utter -- add corresponding labels diff --git a/htk_io/tools/nerv_to_tnet.lua b/htk_io/tools/nerv_to_tnet.lua new file mode 100644 index 0000000..c0ac76b --- /dev/null +++ b/htk_io/tools/nerv_to_tnet.lua @@ -0,0 +1,60 @@ +-- usage: nerv config_file nerv_param_input tnet_output + +dofile(arg[1]) +param_repo = nerv.ParamRepo() +param_repo:import({arg[2], gconf.initialized_param[2]}, nil, gconf) +layer_repo = make_layer_repo(param_repo) +f = assert(io.open(arg[3], "w")) + +function print_tnet_matrix(cumat) + local strs = {} + collectgarbage() + if cumat:nrow() == 1 then + local mat = nerv.MMatrixFloat(1, cumat:ncol()) + cumat:copy_toh(mat) + table.insert(strs, string.format("v %d\n", mat:ncol())) + for j = 0, mat:ncol() - 1 do + table.insert(strs, string.format("%.8f ", mat[0][j])) + end + table.insert(strs, "\n") + f:write(table.concat(strs)) + else + cumat = cumat:trans() + local mat = nerv.MMatrixFloat(cumat:nrow(), cumat:ncol()) + cumat:copy_toh(mat) + table.insert(strs, string.format("m %d %d\n", mat:nrow(), mat:ncol())) + for i = 0, mat:nrow() - 1 do + local row = mat[i] + for j = 0, mat:ncol() - 1 do + table.insert(strs, string.format("%.8f ", row[j])) + end + table.insert(strs, "\n") + f:write(table.concat(strs)) + strs = {} + end + end +end +local lnames = {"affine0", "sigmoid0", + "affine1", "sigmoid1", + "affine2", "sigmoid2", + "affine3", "sigmoid3", + "affine4", "sigmoid4", + "affine5", "sigmoid5", + "affine6", "sigmoid6", + "affine7", "ce_crit"} +for i, name in ipairs(lnames) do + local layer = layer_repo:get_layer(name) + local layer_type = layer.__typename + if layer_type == "nerv.AffineLayer" then + f:write(string.format("<biasedlinearity> %d %d\n", layer.dim_out[1], layer.dim_in[1])) + print_tnet_matrix(layer.ltp.trans) + print_tnet_matrix(layer.bp.trans) + elseif layer_type == "nerv.SigmoidLayer" then + f:write(string.format("<sigmoid> %d %d\n", layer.dim_out[1], layer.dim_in[1])) + elseif layer_type == "nerv.SoftmaxCELayer" then + f:write(string.format("<softmax> %d %d\n", layer.dim_in[1], layer.dim_in[1])) + else + nerv.error("unknown layer type %s", layer_type) + end +end +f:close() diff --git a/htk_io/tools/tnet_to_nerv.c b/htk_io/tools/tnet_to_nerv.c index f781236..5774819 100644 --- a/htk_io/tools/tnet_to_nerv.c +++ b/htk_io/tools/tnet_to_nerv.c @@ -41,7 +41,7 @@ int main() { fprintf(fout, "%16d", 0); fprintf(fout, "{type=\"nerv.BiasParam\",id=\"affine%d_bp\"}\n", cnt); - fprintf(fout, "1 %d\n", nrow, ncol); + fprintf(fout, "1 %d\n", ncol); for (j = 0; j < ncol; j++) fprintf(fout, "%.8f ", mat[0][j]); fprintf(fout, "\n"); diff --git a/htk_io/tools/tnet_to_nerv.cpp b/htk_io/tools/tnet_to_nerv.cpp index bbfddcf..a779a25 100644 --- a/htk_io/tools/tnet_to_nerv.cpp +++ b/htk_io/tools/tnet_to_nerv.cpp @@ -2,9 +2,10 @@ #include <fstream> #include <string> #include <cstring> +#include <cstdlib> char token[1024]; char output[1024]; -double mat[4096][4096]; +double **mat; int main(int argc, char **argv) { std::ofstream fout; fout.open(argv[1]); @@ -18,6 +19,9 @@ int main(int argc, char **argv) { scanf("%d %d", &ncol, &nrow); scanf("%s %d %d", token, &ncol, &nrow); printf("%d %d\n", nrow, ncol); + mat = (double **)malloc(nrow * sizeof(double *)); + for (i = 0; i < nrow; i++) + mat[i] = (double *)malloc(ncol * sizeof(double)); for (j = 0; j < ncol; j++) for (i = 0; i < nrow; i++) scanf("%lf", mat[i] + j); @@ -34,7 +38,9 @@ int main(int argc, char **argv) { for (j = 0; j < ncol; j++) fout << mat[i][j] << " "; fout << std::endl; + free(mat[i]); } + free(mat); long length = fout.tellp() - base; fout.seekp(base); sprintf(output, "[%13lu]\n", length); diff --git a/kaldi_io/example/swb_baseline_basic.lua b/kaldi_io/example/swb_baseline_basic.lua new file mode 100644 index 0000000..e6c8145 --- /dev/null +++ b/kaldi_io/example/swb_baseline_basic.lua @@ -0,0 +1,157 @@ +require 'kaldi_io' +gconf = {lrate = 0.8, wcost = 1e-6, momentum = 0.9, + cumat_type = nerv.CuMatrixFloat, + mmat_type = nerv.MMatrixFloat, + frm_ext = 5, + tr_rspecifier = "ark:/slfs6/users/ymz09/kaldi/src/featbin/copy-feats scp:/slfs6/users/ymz09/swb_ivec/train_bp.scp ark:- |", + cv_rspecifier = "ark:/slfs6/users/ymz09/kaldi/src/featbin/copy-feats scp:/slfs6/users/ymz09/swb_ivec/train_cv.scp ark:- |", + initialized_param = {"/slfs6/users/ymz09/swb_ivec/swb_init.nerv", + "/slfs6/users/ymz09/swb_ivec/swb_global_transf.nerv"}, + debug = false} + +function make_sublayer_repo(param_repo) + return nerv.LayerRepo( + { + -- global transf + ["nerv.BiasLayer"] = + { + blayer1 = {{bias = "bias1"}, {dim_in = {429}, dim_out = {429}}}, + blayer2 = {{bias = "bias2"}, {dim_in = {429}, dim_out = {429}}} + }, + ["nerv.WindowLayer"] = + { + wlayer1 = {{window = "window1"}, {dim_in = {429}, dim_out = {429}}}, + wlayer2 = {{window = "window2"}, {dim_in = {429}, dim_out = {429}}} + }, + -- biased linearity + ["nerv.AffineLayer"] = + { + affine0 = {{ltp = "affine0_ltp", bp = "affine0_bp"}, + {dim_in = {429}, dim_out = {2048}}}, + affine1 = {{ltp = "affine1_ltp", bp = "affine1_bp"}, + {dim_in = {2048}, dim_out = {2048}}}, + affine2 = {{ltp = "affine2_ltp", bp = "affine2_bp"}, + {dim_in = {2048}, dim_out = {2048}}}, + affine3 = {{ltp = "affine3_ltp", bp = "affine3_bp"}, + {dim_in = {2048}, dim_out = {2048}}}, + affine4 = {{ltp = "affine4_ltp", bp = "affine4_bp"}, + {dim_in = {2048}, dim_out = {2048}}}, + affine5 = {{ltp = "affine5_ltp", bp = "affine5_bp"}, + {dim_in = {2048}, dim_out = {2048}}}, + affine6 = {{ltp = "affine6_ltp", bp = "affine6_bp"}, + {dim_in = {2048}, dim_out = {2048}}}, + affine7 = {{ltp = "affine7_ltp", bp = "affine7_bp"}, + {dim_in = {2048}, dim_out = {3001}}} + }, + ["nerv.SigmoidLayer"] = + { + sigmoid0 = {{}, {dim_in = {2048}, dim_out = {2048}}}, + sigmoid1 = {{}, {dim_in = {2048}, dim_out = {2048}}}, + sigmoid2 = {{}, {dim_in = {2048}, dim_out = {2048}}}, + sigmoid3 = {{}, {dim_in = {2048}, dim_out = {2048}}}, + sigmoid4 = {{}, {dim_in = {2048}, dim_out = {2048}}}, + sigmoid5 = {{}, {dim_in = {2048}, dim_out = {2048}}}, + sigmoid6 = {{}, {dim_in = {2048}, dim_out = {2048}}} + }, + ["nerv.SoftmaxCELayer"] = + { + ce_crit = {{}, {dim_in = {3001, 1}, dim_out = {1}, compressed = true}} + } + }, param_repo, gconf) +end + +function make_layer_repo(sublayer_repo, param_repo) + return nerv.LayerRepo( + { + ["nerv.DAGLayer"] = + { + global_transf = {{}, { + dim_in = {429}, dim_out = {429}, + sub_layers = sublayer_repo, + connections = { + ["<input>[1]"] = "blayer1[1]", + ["blayer1[1]"] = "wlayer1[1]", + ["wlayer1[1]"] = "blayer2[1]", + ["blayer2[1]"] = "wlayer2[1]", + ["wlayer2[1]"] = "<output>[1]" + } + }}, + main = {{}, { + dim_in = {429, 1}, dim_out = {1}, + sub_layers = sublayer_repo, + connections = { + ["<input>[1]"] = "affine0[1]", + ["affine0[1]"] = "sigmoid0[1]", + ["sigmoid0[1]"] = "affine1[1]", + ["affine1[1]"] = "sigmoid1[1]", + ["sigmoid1[1]"] = "affine2[1]", + ["affine2[1]"] = "sigmoid2[1]", + ["sigmoid2[1]"] = "affine3[1]", + ["affine3[1]"] = "sigmoid3[1]", + ["sigmoid3[1]"] = "affine4[1]", + ["affine4[1]"] = "sigmoid4[1]", + ["sigmoid4[1]"] = "affine5[1]", + ["affine5[1]"] = "sigmoid5[1]", + ["sigmoid5[1]"] = "affine6[1]", + ["affine6[1]"] = "sigmoid6[1]", + ["sigmoid6[1]"] = "affine7[1]", + ["affine7[1]"] = "ce_crit[1]", + ["<input>[2]"] = "ce_crit[2]", + ["ce_crit[1]"] = "<output>[1]" + } + }} + } + }, param_repo, gconf) +end + +function get_network(layer_repo) + return layer_repo:get_layer("main") +end + +function make_readers(feature_rspecifier, layer_repo) + return { + {reader = nerv.KaldiReader(gconf, + { + id = "main_scp", + feature_rspecifier = feature_rspecifier, + frm_ext = gconf.frm_ext, + mlfs = { + phone_state = { + targets_rspecifier = "ark:/slfs6/users/ymz09/kaldi/src/bin/ali-to-pdf /slfs6/users/ymz09/swb_ivec/final.mdl \"ark:gunzip -c /slfs6/users/ymz09/swb_ivec/ali.*.gz |\" ark:- | /slfs6/users/ymz09/kaldi/src/bin/ali-to-post ark:- ark:- |", + format = "map" + } + }, + global_transf = layer_repo:get_layer("global_transf") + }), + data = {main_scp = 429, phone_state = 1}} + } +end + +function make_buffer(readers) + return nerv.SGDBuffer(gconf, + { + buffer_size = gconf.buffer_size, + randomize = gconf.randomize, + readers = readers + }) +end + +function get_input_order() + return {"main_scp", "phone_state"} +end + +function get_accuracy(sublayer_repo) + local ce_crit = sublayer_repo:get_layer("ce_crit") + return ce_crit.total_correct / ce_crit.total_frames * 100 +end + +function print_stat(sublayer_repo) + local ce_crit = sublayer_repo:get_layer("ce_crit") + nerv.info("*** training stat begin ***") + nerv.printf("cross entropy:\t\t%.8f\n", ce_crit.total_ce) + nerv.printf("correct:\t\t%d\n", ce_crit.total_correct) + nerv.printf("frames:\t\t\t%d\n", ce_crit.total_frames) + nerv.printf("err/frm:\t\t%.8f\n", ce_crit.total_ce / ce_crit.total_frames) + nerv.printf("accuracy:\t\t%.3f%%\n", get_accuracy(sublayer_repo)) + nerv.info("*** training stat end ***") +end diff --git a/kaldi_io/kaldi.mk b/kaldi_io/kaldi.mk new file mode 100644 index 0000000..4a397f0 --- /dev/null +++ b/kaldi_io/kaldi.mk @@ -0,0 +1,70 @@ +# This file was generated using the following command: +# ./configure + +# Rules that enable valgrind debugging ("make valgrind") + +valgrind: .valgrind + +.valgrind: + echo -n > valgrind.out + for x in $(TESTFILES); do echo $$x>>valgrind.out; valgrind ./$$x >/dev/null 2>> valgrind.out; done + ! ( grep 'ERROR SUMMARY' valgrind.out | grep -v '0 errors' ) + ! ( grep 'definitely lost' valgrind.out | grep -v -w 0 ) + rm valgrind.out + touch .valgrind + + +CONFIGURE_VERSION := 2 +OPENFSTLIBS = -L/slwork/users/wd007/src/kaldi/tools/openfst/lib -lfst +OPENFSTLDFLAGS = -Wl,-rpath=/slwork/users/wd007/src/kaldi/tools/openfst/lib +FSTROOT = /slwork/users/wd007/src/kaldi/tools/openfst +ATLASINC = /slwork/users/wd007/src/kaldi/tools/ATLAS/include +ATLASLIBS = -L/usr/lib -llapack -lcblas -latlas -lf77blas +# You have to make sure ATLASLIBS is set... + +ifndef FSTROOT +$(error FSTROOT not defined.) +endif + +ifndef ATLASINC +$(error ATLASINC not defined.) +endif + +ifndef ATLASLIBS +$(error ATLASLIBS not defined.) +endif + + +CXXFLAGS = -msse -msse2 -Wall -I.. \ + -fPIC \ + -DKALDI_DOUBLEPRECISION=0 -DHAVE_POSIX_MEMALIGN \ + -Wno-sign-compare -Wno-unused-local-typedefs -Winit-self \ + -DHAVE_EXECINFO_H=1 -rdynamic -DHAVE_CXXABI_H \ + -DHAVE_ATLAS -I$(ATLASINC) \ + -I$(FSTROOT)/include \ + $(EXTRA_CXXFLAGS) \ + -g # -O0 -DKALDI_PARANOID + +ifeq ($(KALDI_FLAVOR), dynamic) +CXXFLAGS += -fPIC +endif + +LDFLAGS = -rdynamic $(OPENFSTLDFLAGS) +LDLIBS = $(EXTRA_LDLIBS) $(OPENFSTLIBS) $(ATLASLIBS) -lm -lpthread -ldl +CC = g++ +CXX = g++ +AR = ar +AS = as +RANLIB = ranlib + +#Next section enables CUDA for compilation +CUDA = true +CUDATKDIR = /usr/local/cuda + +CUDA_INCLUDE= -I$(CUDATKDIR)/include +CUDA_FLAGS = -g -Xcompiler -fPIC --verbose --machine 64 -DHAVE_CUDA + +CXXFLAGS += -DHAVE_CUDA -I$(CUDATKDIR)/include +CUDA_LDFLAGS += -L$(CUDATKDIR)/lib64 -Wl,-rpath,$(CUDATKDIR)/lib64 +CUDA_LDLIBS += -lcublas -lcudart #LDLIBS : The libs are loaded later than static libs in implicit rule + |