Real optim state hidden because not passed in

src/optim/init.lua

@@ -1,10 +1,11 @@
 local util = require 'autograd.util'
 
 local function wrap(optimfn)
-   return function(fn, state, params)
-      local states = { }
+   return function(fn, config, state, params)
+      local configs, states = { }, { }
       local flatParams = util.sortedFlatten(params)
       for i = 1, #flatParams do
+         configs[i] = util.deepCopy(config)
          states[i] = util.deepCopy(state)
       end
       return function(...)
@@ -15,10 +16,10 @@ local function wrap(optimfn)
             local grad = flatGrads[i]
             optimfn(function()
                return loss, grad
-            end, flatParams[i], states[i])
+            end, flatParams[i], configs[i], states[i])
          end
          return table.unpack(out)
-      end, states
+      end, configs, states
    end
 end
 
@@ -28,4 +29,4 @@ for k, v in pairs(require 'optim') do
    opt[k] = wrap(v)
 end
 
-return opt
+return opt

test/test.lua

@@ -1530,10 +1530,10 @@ local tests = {
          end
       end
 
-      local state = { learningRate = learningRate }
+      local config, state = { learningRate = learningRate }, { }
       local loss3
       for e=1, 5 do
-         local optimfn, states = autograd.optim.sgd(df, state, params3)
+         local optimfn, configs, states = autograd.optim.sgd(df, config, state, params3)
          loss3 = 0
          for i=1,nData do
             local grads, loss = optimfn(xs:narrow(1, i, 1), ys:narrow(1, i, 1))
@@ -1583,7 +1583,7 @@ local tests = {
       local g = autograd(f, {optimize = true})
 
       -- FAILS FOR OTHER OPTIMIZERS AS WELL
-      local optimfn, states = autograd.optim.sgd(g, {learningRate=1e-2}, params)
+      local optimfn, configs, states = autograd.optim.sgd(g, {learningRate=1e-2}, {}, params)
 
       for i=1,3 do
          -- Get images in BHWD format, labels in one-hot format: