move global_transf to asr_trainer.lua

6 files changed, 56 insertions, 12 deletions

diff --git a/nerv/examples/asr_trainer.lua b/nerv/examples/asr_trainer.lua
index dcadfa3..5a50542 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 +25,29 @@ 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,
+                                        gconf.frm_trim or 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 604fa07..f9d281c 100644
--- a/nerv/io/sgd_buffer.lua
+++ b/nerv/io/sgd_buffer.lua
@@ -91,7 +91,7 @@ function SGDBuffer:get_data()
         if not self:saturate() then
             return nil -- the remaining data cannot build a batch
         end
-        nerv.info("%.3fs to fill the buffer\n", os.clock() - t)
+        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/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