.

10 files changed, 65 insertions, 24 deletions

diff --git a/nerv/Makefile b/nerv/Makefile
index 7ed140d..b5d26bd 100644
--- a/nerv/Makefile
+++ b/nerv/Makefile
@@ -36,8 +36,8 @@ LUA_LIBS := matrix/init.lua io/init.lua init.lua \
 			io/sgd_buffer.lua 
 
 INCLUDE := -I $(LUA_INCDIR) -DLUA_USE_APICHECK
-#CUDA_BASE := /usr/local/cuda-6.5
-CUDA_BASE := /usr/local/cuda-5.0
+#CUDA_BASE := /usr/local/cuda-7.0
+CUDA_BASE := /usr/local/cuda
 CUDA_INCLUDE := -I $(CUDA_BASE)/include/
 INCLUDE += $(CUDA_INCLUDE)
 
diff --git a/nerv/examples/asr_trainer.lua b/nerv/examples/asr_trainer.lua
index dcadfa3..69cfeed 100644
--- a/nerv/examples/asr_trainer.lua
+++ b/nerv/examples/asr_trainer.lua
@@ -3,6 +3,7 @@ function build_trainer(ifname)
     param_repo:import(ifname, nil, gconf)
     local layer_repo = make_layer_repo(param_repo)
     local network = get_network(layer_repo)
+    local global_transf = get_global_transf(layer_repo)
     local input_order = get_input_order()
     local iterative_trainer = function (prefix, scp_file, bp)
         gconf.randomize = bp
@@ -24,15 +25,28 @@ function build_trainer(ifname)
                 -- break
             end
             local input = {}
---            if gconf.cnt == 100 then break end
-            for i, id in ipairs(input_order) do
+--            if gconf.cnt == 1000 then break end
+            for i, e in ipairs(input_order) do
+                local id = e.id
                 if data[id] == nil then
                     nerv.error("input data %s not found", id)
                 end
-                table.insert(input, data[id])
+                local transformed
+                if e.global_transf then
+                    transformed = nerv.speech_utils.global_transf(data[id],
+                                        global_transf,
+                                        gconf.frm_ext or 0, 0,
+                                        gconf)
+                else
+                    transformed = data[id]
+                end
+                table.insert(input, transformed)
             end
             local output = {nerv.CuMatrixFloat(gconf.batch_size, 1)}
-            err_output = {input[1]:create()}
+            err_output = {}
+            for i = 1, #input do
+                table.insert(err_output, input[i]:create())
+            end
             network:propagate(input, output)
             if bp then
                 network:back_propagate(err_input, err_output, input, output)
diff --git a/nerv/examples/swb_baseline.lua b/nerv/examples/swb_baseline.lua
index 0e9f897..bbc6467 100644
--- a/nerv/examples/swb_baseline.lua
+++ b/nerv/examples/swb_baseline.lua
@@ -3,6 +3,7 @@ gconf = {lrate = 0.8, wcost = 1e-6, momentum = 0.9,
         cumat_type = nerv.CuMatrixFloat,
         mmat_type = nerv.MMatrixFloat,
         frm_ext = 5,
+        frm_trim = 5,
         tr_scp = "/slfs1/users/mfy43/swb_ivec/train_bp.scp",
         cv_scp = "/slfs1/users/mfy43/swb_ivec/train_cv.scp",
         htk_conf = "/slfs1/users/mfy43/swb_ivec/plp_0_d_a.conf",
@@ -161,8 +162,7 @@ function make_readers(scp_file, layer_repo)
                                 dir = "*/",
                                 ext = "lab"
                             }
-                        },
-                        global_transf = layer_repo:get_layer("global_transf")
+                        }
                     }),
                 data = {main_scp = 429, phone_state = 1}}
             }
@@ -178,7 +178,8 @@ function make_buffer(readers)
 end
 
 function get_input_order()
-    return {"main_scp", "phone_state"}
+    return {{id = "main_scp", global_transf = true},
+            {id = "phone_state"}}
 end
 
 function get_accuracy(layer_repo)
diff --git a/nerv/examples/swb_baseline_basic.lua b/nerv/examples/swb_baseline_basic.lua
index c47ec3e..71f04a3 100644
--- a/nerv/examples/swb_baseline_basic.lua
+++ b/nerv/examples/swb_baseline_basic.lua
@@ -3,6 +3,7 @@ gconf = {lrate = 0.8, wcost = 1e-6, momentum = 0.9,
         cumat_type = nerv.CuMatrixFloat,
         mmat_type = nerv.MMatrixFloat,
         frm_ext = 5,
+        frm_trim = 5,
         tr_scp = "/slfs1/users/mfy43/swb_ivec/train_bp.scp",
         cv_scp = "/slfs1/users/mfy43/swb_ivec/train_cv.scp",
         htk_conf = "/slfs1/users/mfy43/swb_ivec/plp_0_d_a.conf",
@@ -124,8 +125,7 @@ function make_readers(scp_file, layer_repo)
                                 dir = "*/",
                                 ext = "lab"
                             }
-                        },
-                        global_transf = layer_repo:get_layer("global_transf")
+                        }
                     }),
                 data = {main_scp = 429, phone_state = 1}}
             }
@@ -141,7 +141,8 @@ function make_buffer(readers)
 end
 
 function get_input_order()
-    return {"main_scp", "phone_state"}
+    return {{id = "main_scp", global_transf = true},
+            {id = "phone_state"}}
 end
 
 function get_accuracy(layer_repo)
diff --git a/nerv/io/sgd_buffer.lua b/nerv/io/sgd_buffer.lua
index f4f7dfe..f9d281c 100644
--- a/nerv/io/sgd_buffer.lua
+++ b/nerv/io/sgd_buffer.lua
@@ -41,7 +41,7 @@ function SGDBuffer:saturate()
                 buff.data:copy_from(buff.leftover, 0, lrow)
                 buff.leftover = nil
             end
-            nerv.printf("leftover: %d\n", lrow)
+            nerv.printf("buffer leftover: %d\n", lrow)
             reader.tail = lrow
             reader.has_leftover = false
         end
@@ -87,9 +87,11 @@ end
 function SGDBuffer:get_data()
     local batch_size = self.gconf.batch_size
     if self.head >= self.tail then -- buffer is empty
+        local t = os.clock()
         if not self:saturate() then
             return nil -- the remaining data cannot build a batch
         end
+        nerv.info("%.3fs to fill the buffer", os.clock() - t)
     end
     if self.head + batch_size > self.tail then
         return nil -- the remaining data cannot build a batch
diff --git a/nerv/layer/mse.lua b/nerv/layer/mse.lua
index 9a97add..2516998 100644
--- a/nerv/layer/mse.lua
+++ b/nerv/layer/mse.lua
@@ -34,7 +34,7 @@ function MSELayer:propagate(input, output)
     if output[1] ~= nil then
         output[1]:copy_fromd(mse_sum)
     end
-    self.total_mse = self.total_mse + mse_sum:colsum()[0]
+    self.total_mse = self.total_mse + mse_sum:colsum()[0][0]
     self.total_frames = self.total_frames + mse_sum:nrow()
 end
 
diff --git a/nerv/lib/matrix/generic/cumatrix.c b/nerv/lib/matrix/generic/cumatrix.c
index 40a0030..2cb3563 100644
--- a/nerv/lib/matrix/generic/cumatrix.c
+++ b/nerv/lib/matrix/generic/cumatrix.c
@@ -321,6 +321,7 @@ void nerv_matrix_(copy_rows_fromh_by_idx)(Matrix *a, const Matrix *b,
         NERV_EXIT_STATUS(status, MAT_IDX_VECTOR_EXP, 0);
     if (a->ncol != b->ncol)
         NERV_EXIT_STATUS(status, MAT_MISMATCH_DIM, 0);
+    PROFILE_START
     cudaStream_t *streams = (cudaStream_t*)malloc(sizeof(cudaStream_t) * nrow);
     for (i = 0; i < nrow; i++)
     {
@@ -339,6 +340,7 @@ void nerv_matrix_(copy_rows_fromh_by_idx)(Matrix *a, const Matrix *b,
         CUDA_SAFE_CALL(cudaStreamDestroy(streams[i]), status);
     }
     free(streams);
+    PROFILE_STOP
     NERV_SET_STATUS(status, NERV_NORMAL, 0);
 }
 
diff --git a/nerv/matrix/generic/cumatrix.c b/nerv/matrix/generic/cumatrix.c
index 4bdf5f0..ab7f7c4 100644
--- a/nerv/matrix/generic/cumatrix.c
+++ b/nerv/matrix/generic/cumatrix.c
@@ -43,15 +43,6 @@ static int nerv_matrix_(lua_mul)(lua_State *L) {
     return 0;
 }
 
-static int nerv_matrix_(lua_create)(lua_State *L) {
-    Status status;
-    Matrix *a = luaT_checkudata(L, 1, nerv_matrix_(tname));
-    Matrix *b = nerv_matrix_(create)(a->nrow, a->ncol, &status);
-    NERV_LUA_CHECK_STATUS(L, status);
-    luaT_pushudata(L, b, nerv_matrix_(tname));
-    return 1;
-}
-
 static int nerv_matrix_(lua_sigmoid)(lua_State *L) {
     Status status;
     Matrix *a = luaT_checkudata(L, 1, nerv_matrix_(tname));
@@ -289,7 +280,6 @@ static int nerv_matrix_(lua_scale_rows_by_row)(lua_State *L) {
 }
 
 static const luaL_Reg nerv_matrix_(extra_methods)[] = {
-    {"create", nerv_matrix_(lua_create)},
     {"colsum", nerv_matrix_(lua_colsum)},
     {"colsame", nerv_matrix_(lua_colsame)},
     {"rowsum", nerv_matrix_(lua_rowsum)},
diff --git a/nerv/matrix/init.lua b/nerv/matrix/init.lua
index f230e9f..1091d7e 100644
--- a/nerv/matrix/init.lua
+++ b/nerv/matrix/init.lua
@@ -45,6 +45,10 @@ function nerv.Matrix:generate(gen)
     end
 end
 
+function nerv.Matrix:create(nrow, ncol)
+    return self.__constructor(nrow or self:nrow(), ncol or self:ncol())
+end
+
 nerv.MMatrixInt.fmt = "%d "
 
 function nerv.CuMatrix:__add__(b)
diff --git a/nerv/nn/layer_dag.lua b/nerv/nn/layer_dag.lua
index e9d4d86..25297c2 100644
--- a/nerv/nn/layer_dag.lua
+++ b/nerv/nn/layer_dag.lua
@@ -254,3 +254,30 @@ function DAGLayer:get_params()
     end
     return nerv.ParamRepo.merge(param_repos)
 end
+
+DAGLayer.PORT_TYPES = {
+    INPUT = {},
+    OUTPUT = {},
+    ERR_INPUT = {},
+    ERR_OUTPUT = {}
+}
+
+function DAGLayer:get_intermediate(id, port_type)
+    if id == "<input>" or id == "<output>" then
+        nerv.error("an actual real layer id is expected")
+    end
+    local layer = layers[id]
+    if layer == nil then
+        nerv.error("layer id %s not found", id)
+    end
+    if port_type == DAGLayer.PORT_TYPES.INPUT then
+        return layer.inputs
+    elseif port_type == DAGLayer.PORT_TYPES.OUTPUT then
+        return layer.outputs
+    elseif port_type == DAGLayer.PORT_TYPES.ERR_INPUT then
+        return layer.err_inputs
+    elseif port_type == DAGLayer.PORT_TYPES.ERR_OUTPUT then
+        return layer.err_outputs
+    end
+    nerv.error("unrecognized port type")
+end