1 files changed, 67 insertions, 0 deletions
diff --git a/nerv/examples/lmptb/rnn/layers/gate_fff.lua b/nerv/examples/lmptb/rnn/layers/gate_fff.lua
new file mode 100644
index 0000000..74e19ce
--- /dev/null
+++ b/nerv/examples/lmptb/rnn/layers/gate_fff.lua
@@ -0,0 +1,67 @@
+local GateFFFLayer = nerv.class('nerv.GateFFFLayer', 'nerv.Layer')
+
+function AffineLayer:__init(id, global_conf, layer_conf)
+    self.id = id
+    self.ltp = layer_conf.ltp
+    self.bp = layer_conf.bp
+    self.dim_in = layer_conf.dim_in
+    self.dim_out = layer_conf.dim_out
+    self.gconf = global_conf
+    self:check_dim_len(1, 1) -- exactly one input and one output
+end
+
+function AffineLayer:init(batch_size)
+    if self.ltp.trans:ncol() ~= self.bp.trans:ncol() then
+        nerv.error("mismatching dimensions of linear transform and bias paramter")
+    end
+    if self.dim_in[1] ~= self.ltp.trans:nrow() then
+        nerv.error("mismatching dimensions of linear transform parameter and input")
+    end
+    if self.dim_out[1] ~= self.ltp.trans:ncol() then
+        nerv.error("mismatching dimensions of linear transform parameter and output")
+    end
+    self.ltp_grad = self.ltp.trans:create()
+    self.ltp:train_init()
+    self.bp:train_init()
+end
+
+function AffineLayer:batch_resize(batch_size)
+    -- do nothing
+end
+
+function AffineLayer:update(bp_err, input, output)
+    if self.direct_update == true then
+        local gconf = self.gconf
+        if gconf.momentum > 0 then
+            self.ltp.correction:mul(input[1], bp_err[1], 1.0, gconf.momentum, 'T', 'N')
+            self.bp.correction:add(self.bp.correction, bp_err[1]:colsum(), gconf.momentum, 1)
+            -- momentum gain
+            local mmt_gain = 1.0 / (1.0 - gconf.momentum)
+            local n = self.gconf.batch_size * mmt_gain
+            -- perform update
+            self.ltp.trans:add(self.ltp.trans, self.ltp.correction, 1.0 - gconf.lrate * gconf.wcost / gconf.batch_size, - gconf.lrate / n)
+            self.bp.trans:add(self.bp.trans, self.bp.correction, 1.0 - gconf.lrate * gconf.wcost / gconf.batch_size, - gconf.lrate / n)
+        else
+            self.ltp.trans:mul(input[1], bp_err[1], - gconf.lrate / gconf.batch_size, 1.0 - gconf.lrate * gconf.wcost / gconf.batch_size, 'T', 'N')
+            self.bp.trans:add(self.bp.trans, bp_err[1]:colsum(), 1.0 - gconf.lrate * gconf.wcost / gconf.batch_size, - gconf.lrate / gconf.batch_size)
+        end
+    else
+        self.ltp:update_by_err_input(bp_err[1], input[1])
+        self.bp:update_by_gradient(bp_err[1]:colsum())
+    end
+end
+
+function AffineLayer:propagate(input, output)
+    -- apply linear transform
+    output[1]:mul(input[1], self.ltp.trans, 1.0, 0.0, 'N', 'N')
+    -- add bias
+    output[1]:add_row(self.bp.trans, 1.0)
+end
+
+function AffineLayer:back_propagate(bp_err, next_bp_err, input, output)
+    next_bp_err[1]:mul(bp_err[1], self.ltp.trans, 1.0, 0.0, 'N', 'T')
+end
+
+function AffineLayer:get_params()
+    return nerv.ParamRepo({self.ltp, self.bp})
+end