trying to test softmax_ce_t

5 files changed, 131 insertions, 6 deletions

diff --git a/nerv/examples/lmptb/lmptb/lmseqreader.lua b/nerv/examples/lmptb/lmptb/lmseqreader.lua
index 41e3903..f7e2539 100644
--- a/nerv/examples/lmptb/lmptb/lmseqreader.lua
+++ b/nerv/examples/lmptb/lmptb/lmseqreader.lua
@@ -1,5 +1,4 @@
 require 'lmptb.lmvocab'
-require 'rnn.tnn'
 
 local LMReader = nerv.class("nerv.LMSeqReader")
 
diff --git a/nerv/examples/lmptb/m-tests/tnn_test.lua b/nerv/examples/lmptb/m-tests/tnn_test.lua
index ddea54c..888ba0f 100644
--- a/nerv/examples/lmptb/m-tests/tnn_test.lua
+++ b/nerv/examples/lmptb/m-tests/tnn_test.lua
@@ -2,8 +2,8 @@ require 'lmptb.lmvocab'
 require 'lmptb.lmfeeder'
 require 'lmptb.lmutil'
 require 'lmptb.layer.init'
+require 'rnn.init'
 require 'lmptb.lmseqreader'
-require 'rnn.tnn'
 
 --[[global function rename]]--
 printf = nerv.printf
@@ -194,6 +194,7 @@ function lm_process_file(global_conf, fn, tnn, do_train)
             next_log_wcn = next_log_wcn + global_conf.log_w_num
             printf("%s %d words processed %s.\n", global_conf.sche_log_pre, result["rnn"].cn_w, os.date()) 
             printf("\t%s log prob per sample :%f.\n", global_conf.sche_log_pre, result:logp_sample("rnn"))
+            nerv.LMUtil.wait(1)
         end
          
         --[[
@@ -259,17 +260,16 @@ global_conf = {
     nn_act_default = 0, 
 
     hidden_size = 20,
-    chunk_size = 5,
+    chunk_size = 2,
     batch_size = 3, 
     max_iter = 3,
     param_random = function() return (math.random() / 5 - 0.1) end,
-    independent = true,
 
     train_fn = train_fn,
     valid_fn = valid_fn,
     test_fn = test_fn,
     sche_log_pre = "[SCHEDULER]:",
-    log_w_num = 20, --give a message when log_w_num words have been processed
+    log_w_num = 10, --give a message when log_w_num words have been processed
     timer = nerv.Timer()
 }
 
diff --git a/nerv/examples/lmptb/rnn/init.lua b/nerv/examples/lmptb/rnn/init.lua
new file mode 100644
index 0000000..0e08cb6
--- /dev/null
+++ b/nerv/examples/lmptb/rnn/init.lua
@@ -0,0 +1,45 @@
+local Layer = nerv.class('nerv.LayerT')
+
+function Layer:__init(id, global_conf, layer_conf)
+    nerv.error_method_not_implemented()
+end
+
+function Layer:init(batch_size, chunk_size)
+    nerv.error_method_not_implemented()
+end
+
+function Layer:update(bp_err, input, output, t)
+    nerv.error_method_not_implemented()
+end
+
+function Layer:propagate(input, output, t)
+    nerv.error_method_not_implemented()
+end
+
+function Layer:back_propagate(bp_err, next_bp_err, input, output, t)
+    nerv.error_method_not_implemented()
+end
+
+function Layer:check_dim_len(len_in, len_out)
+    local expected_in = #self.dim_in
+    local expected_out = #self.dim_out
+    if len_in > 0 and expected_in ~= len_in then
+        nerv.error("layer %s expects %d inputs, %d given",
+                    self.id, len_in, expected_in)
+    end
+    if len_out > 0 and expected_out ~= len_out then
+        nerv.error("layer %s expects %d outputs, %d given",
+                    self.id, len_out, expected_out)
+    end
+end
+
+function Layer:get_params()
+    nerv.error_method_not_implemented()
+end
+
+function Layer:get_dim()
+    return self.dim_in, self.dim_out
+end
+
+nerv.include('tnn.lua')
+nerv.include('softmax_ce_t.lua')
diff --git a/nerv/examples/lmptb/rnn/softmax_ce_t.lua b/nerv/examples/lmptb/rnn/softmax_ce_t.lua
new file mode 100644
index 0000000..dddb05a
--- /dev/null
+++ b/nerv/examples/lmptb/rnn/softmax_ce_t.lua
@@ -0,0 +1,81 @@
+local SoftmaxCELayer = nerv.class("nerv.SoftmaxCELayerT", "nerv.LayerT")
+
+function SoftmaxCELayer:__init(id, global_conf, layer_conf)
+    self.id = id
+    self.gconf = global_conf
+    self.dim_in = layer_conf.dim_in
+    self.dim_out = layer_conf.dim_out
+    self.compressed = layer_conf.compressed
+    if self.compressed == nil then
+        self.compressed = false
+    end
+    self:check_dim_len(2, -1) -- two inputs: nn output and label
+end
+
+function SoftmaxCELayer:init(batch_size, chunk_size)
+    if not self.compressed and (self.dim_in[1] ~= self.dim_in[2]) then
+        nerv.error("mismatching dimensions of previous network output and labels")
+    end
+    self.total_ce = 0.0
+    self.total_correct = 0
+    self.total_frames = 0
+    self.softmax_t = {}
+    self.ce_t = {}
+    for t = 1, chunk_size do
+        self.softmax_t[t] = self.gconf.cumat_type(batch_size, self.dim_in[1])
+        self.ce_t[t] = self.gconf.cumat_type(batch_size, self.dim_in[1])
+    end
+end
+
+function SoftmaxCELayer:batch_resize(batch_size)
+    for t = 1, chunk_size do
+        if self.softmax_t[t]:nrow() ~= batch_resize then
+            self.softmax_t[t] = self.gconf.cumat_type(batch_size, self.dim_in[1])
+            self.ce_t[t] = self.gconf.cumat_type(batch_size, self.dim_in[1])
+        end
+    end
+end
+
+function SoftmaxCELayer:update(bp_err, input, output, t)
+    -- no params, therefore do nothing
+end
+
+function SoftmaxCELayer:propagate(input, output, t)
+    local softmax = self.softmax_t[t]
+    local ce = self.ce_t[t]
+    local classified = softmax:softmax(input[1])
+    local label = input[2]
+    ce:log_elem(softmax)
+    if self.compressed then
+        label = label:decompress(input[1]:ncol())
+    end
+    ce:mul_elem(ce, label)
+    ce = ce:rowsum()
+    if output[1] ~= nil then
+        output[1]:copy_fromd(ce)
+    end
+    -- add total ce
+    self.total_ce = self.total_ce - ce:colsum()[0][0]
+    self.total_frames = self.total_frames + softmax:nrow()
+    -- TODO: add colsame for uncompressed label
+    if self.compressed then
+        self.total_correct = self.total_correct + classified:colsame(input[2])[0][0]
+    end
+end
+
+function SoftmaxCELayer:back_propagate(bp_err, next_bp_err, input, output, t)
+    -- softmax output - label
+    local label = input[2]
+    if self.compressed then
+        label = label:decompress(input[1]:ncol())
+    end
+    local nbe = next_bp_err[1]
+    nbe:add(self.softmax_t[t], label, 1.0, -1.0)
+    if bp_err[1] ~= nil then
+        nbe:scale_rows_by_col(bp_err[1])
+    end
+end
+
+function SoftmaxCELayer:get_params()
+    return nerv.ParamRepo({})
+end
diff --git a/nerv/examples/lmptb/rnn/tnn.lua b/nerv/examples/lmptb/rnn/tnn.lua
index ae9ed7a..8c3963c 100644
--- a/nerv/examples/lmptb/rnn/tnn.lua
+++ b/nerv/examples/lmptb/rnn/tnn.lua
@@ -198,7 +198,7 @@ function TNN:init(batch_size, chunk_size)
             end
         end
         -- initialize sub layers
-        ref.layer:init(batch_size)
+        ref.layer:init(batch_size, chunk_size)
     end
 
     local flags_now = {}