Update.

diff --git a/profiler.lua b/profiler.lua
index 91b0915..49eaed4 100755 (executable)
--- a/profiler.lua
+++ b/profiler.lua
@@ -14,12 +14,19 @@ function profiler.decor(model, functionsToDecorate)
       }
 
    for _, name in pairs(functionsToDecorate) do
-      model.orig = model.orig or {}
       model.timings = 0
 
-      if model[name] and not model.orig[name] then
-         model.orig[name] = model[name]
-         model[name] = function(self, ...)
+      local functionTable = model
+
+      if not rawget(functionTable, name) then
+         functionTable = getmetatable(model)
+      end
+
+      if functionTable[name] and not (functionTable.orig and functionTable.orig[name]) then
+         print('Profiler decoring ' .. functionTable.__typename .. '.' .. name)
+         functionTable.orig = functionTable.orig or {}
+         functionTable.orig[name] = functionTable[name]
+         functionTable[name] = function(self, ...)
             local startTime = sys.clock()
             local result = { self.orig[name](self, unpack({...})) }
             local endTime = sys.clock()
@@ -38,12 +45,19 @@ function profiler.decor(model, functionsToDecorate)
 
 end
 
-function profiler.print(model)
+function profiler.print(model, nbSamples)
    print('----------------------------------------------------------------------')
    print(model)
-   print(string.format('TIMING %.02fs', model.timings))
+   if nbSamples then
+      print(string.format('acc_time %.02fs (%.1ems/sample)', model.timings, 1000 * model.timings / nbSamples))
+   else
+      print(string.format('acc_time %.02fs', model.timings))
+   end
+
    if torch.isTypeOf(model, nn.Container) then
-      model:applyToModules(profiler.print)
+      for _, m in ipairs(model.modules) do
+         profiler.print(m, nbSamples)
+      end
    end
 end
 

diff --git a/test-profiler.lua b/test-profiler.lua
index b394a33..44bbee1 100755 (executable)
--- a/test-profiler.lua
+++ b/test-profiler.lua
@@ -12,14 +12,36 @@ model:add(nn.Linear(1000, 100))
 
 profiler.decor(model)
 
-for k = 1, 10 do
-   local input = torch.Tensor(1000, 1000):uniform(-1, 1)
-   local target = torch.Tensor(input:size(1), 100):uniform()
-   local criterion = nn.MSECriterion()
+local input = torch.Tensor(1000, 1000)
+local target = torch.Tensor(input:size(1), 100)
+local criterion = nn.MSECriterion()
+
+local nbSamples = 0
+local modelTime = 0
+local dataTime = 0
+
+for k = 1, 5 do
+   local t1 = sys.clock()
+   input:uniform(-1, 1)
+   target:uniform()
+
+   local t2 = sys.clock()
+
    local output = model:forward(input)
    local loss = criterion:forward(output, target)
    local dloss = criterion:backward(output, target)
    model:backward(input, dloss)
+
+   local t3 = sys.clock()
+
+   dataTime = dataTime + (t2 - t1)
+   modelTime = modelTime + (t3 - t2)
+
+   nbSamples = nbSamples + input:size(1)
 end
 
-profiler.print(model)
+profiler.print(model, nbSamples)
+
+print('----------------------------------------------------------------------')
+print(string.format('Total model time %.02fs', modelTime))
+print(string.format('Total data time %.02fs', dataTime))