Fixed tabs.

1 files changed, 26 insertions, 26 deletions

diff --git a/BatchOptimization.lua b/BatchOptimization.lua
index 27309f2..3143a3e 100644
--- a/BatchOptimization.lua
+++ b/BatchOptimization.lua
@@ -8,15 +8,15 @@ local Batch,parent = torch.class('nn.BatchOptimization', 'nn.Optimization')
 function Batch:__init(...)
    parent.__init(self)
    xlua.unpack_class(self, {...},
-		     'BatchOptimization', nil,
-		     {arg='module', type='nn.Module', help='a module to train', req=true},
-		     {arg='criterion', type='nn.Criterion', 
-		      help='a criterion to estimate the error', req=true},
-		     {arg='parallelize', type='number', 
-		      help='parallelize onto N cores (experimental!)', default=1},
-		     {arg='verbose', type='number', 
-		      help='verbose level during training [0-2]', default=0}
-		  )
+                     'BatchOptimization', nil,
+                     {arg='module', type='nn.Module', help='a module to train', req=true},
+                     {arg='criterion', type='nn.Criterion',
+                      help='a criterion to estimate the error', req=true},
+                     {arg='parallelize', type='number',
+                      help='parallelize onto N cores (experimental!)', default=1},
+                     {arg='verbose', type='number',
+                      help='verbose level during training [0-2]', default=0}
+                  )
    self.parameters = nnx.flattenParameters(nnx.getParameters(self.module))
    self.gradParameters = nnx.flattenParameters(nnx.getGradParameters(self.module))
    self.evalCounter = 0
@@ -45,7 +45,7 @@ function Batch:forward_sequential(inputs, targets, options)
       = function()
            -- verbose
            if self.verbose >= 2 then
-              print('<BatchOptimization> evaluating f(X) + df/dX') 
+              print('<BatchOptimization> evaluating f(X) + df/dX')
            end
            local _t_ = sys.clock()
            -- reset gradients
@@ -75,7 +75,7 @@ function Batch:forward_sequential(inputs, targets, options)
            -- normalize gradients
            self.gradParameters:div(#inputs)
            -- verbose
-           if self.verbose >= 2 then 
+           if self.verbose >= 2 then
               print('<BatchOptimization> ' .. self.evalCounter .. 'th evaluation took ' .. (sys.clock() - _t_) .. ' sec')
            end
            -- return average f(X)
@@ -160,17 +160,17 @@ function Batch:forward_mapreduce(inputs, targets, options)
    self.evaluate
       = function()
            -- verbose
-           if self.verbose >= 2 then 
-              print('<BatchOptimization> evaluating f(X) + df/dX') 
+           if self.verbose >= 2 then
+              print('<BatchOptimization> evaluating f(X) + df/dX')
            end
            local _t_ = sys.clock()
            -- do map/reduce
-	   self.evaluate_map()
-	   self.evaluate_reduce()
-	   -- update evaluation counter
+           self.evaluate_map()
+           self.evaluate_reduce()
+           -- update evaluation counter
            self.evalCounter = self.evalCounter + 1
            -- verbose
-           if self.verbose >= 2 then 
+           if self.verbose >= 2 then
               print('<BatchOptimization> ' .. self.evalCounter .. 'th evaluation took ' .. (sys.clock() - _t_) .. ' sec')
            end
            return self.output
@@ -180,12 +180,12 @@ function Batch:forward_mapreduce(inputs, targets, options)
    --      in separate threads
    self.evaluate_map
       = function()
-	   -- transmit new parameters to all workers
+           -- transmit new parameters to all workers
            parallel.children:send(self.parameters)
-	   -- then wait for all workers to return their partial gradParameters + outputs
+           -- then wait for all workers to return their partial gradParameters + outputs
            gradParametersPartial = parallel.children:receive()
-	   outputsPartial = parallel.children:receive()
-	   -- force cleanup
+           outputsPartial = parallel.children:receive()
+           -- force cleanup
            collectgarbage()
         end
 
@@ -233,7 +233,7 @@ function Batch:setup_mapreduce ()
 
    -- (1) define code for workers
    local worker_code = [[
-	 -- require packages
+         -- require packages
          require 'nnx'
 
          -- retrieve module + criterion at startup
@@ -260,11 +260,11 @@ function Batch:setup_mapreduce ()
             if type(inputs) == 'string' and inputs == 'break' then break end
             targets = parallel.parent:receive()
             options = parallel.parent:receive()
-	    -- inner loop: evaluations
+            -- inner loop: evaluations
             while true do
-	       -- receive new set of parameters
+               -- receive new set of parameters
                newParameters = parallel.parent:receive()
-	       
+
                if type(newParameters) == 'string' and newParameters == 'break' then break end
                parameters:copy(newParameters)
 
@@ -274,7 +274,7 @@ function Batch:setup_mapreduce ()
                local f_x = 0
                -- evaluate gradients on inputs for this thread
                for i = 1,#inputs do
-		  -- user hook
+                  -- user hook
                   if prehook then
                      prehook(optimizer, {inputs[i], targets[i], options[i]})
                   end