eeva/tensorflow-haskell
1
0
Fork 0
mirror of https://github.com/tensorflow/haskell.git synced 2024-11-23 03:19:44 +01:00
Code Issues Releases Wiki Activity

Make 'mean' doubly differentiable (#241)

Browse source 
Use stopGradient on shape computations
Add opGrad for StopGradient
This commit is contained in:
Christian Berentsen 2019-04-22 06:46:01 +02:00 committed by fkm3
parent 7316062c10
commit 4a2e46ba57
2 changed files with 22 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

4
tensorflow-ops/src/TensorFlow/Gradient.hs
View file

@ -574,7 +574,7 @@ opGrad "Sum" _ [toT -> x, toT -> indices] [dz] =
grad = reshape dz outputShapeKeptDims grad = reshape dz outputShapeKeptDims
opGrad "Mean" u v@[toT -> x, _] w = opGrad "Mean" u v@[toT -> x, _] w =
[Just $ dz `CoreOps.div` CoreOps.cast factor, Nothing] [Just $ dz `CoreOps.div` (CoreOps.stopGradient $ CoreOps.cast $ factor), Nothing]
where where
[Just dz, Nothing] = opGrad "Sum" u v w [Just dz, Nothing] = opGrad "Sum" u v w
inputShape = shape (x :: Tensor Build a) inputShape = shape (x :: Tensor Build a)
@ -858,6 +858,7 @@ opGrad "Fill" _ _ [dz] = [Nothing, Just $ sum dz rx]
opGrad "ReadVariableOp" _ _ [dz] = [Just $ expr dz] opGrad "ReadVariableOp" _ _ [dz] = [Just $ expr dz]
opGrad "Const" _ _ _ = [Nothing, Nothing] opGrad "Const" _ _ _ = [Nothing, Nothing]
opGrad "StopGradient" _ _ _ = [Nothing]
opGrad "VarHandleOp" _ _ _ = [] opGrad "VarHandleOp" _ _ _ = []
opGrad "Sqrt" _ [toT -> x] [dz] = [Just $ sq' `CoreOps.mul` dz] opGrad "Sqrt" _ [toT -> x] [dz] = [Just $ sq' `CoreOps.mul` dz]
@ -901,6 +902,7 @@ numOutputs o =
"Neg" -> 1 "Neg" -> 1
"Pad" -> 1 "Pad" -> 1
"Placeholder" -> 1 "Placeholder" -> 1
"StopGradient" -> 1
"OneHot" -> 1 "OneHot" -> 1
"ReadVariableOp" -> 1 "ReadVariableOp" -> 1
"RefIdentity" -> 1 "RefIdentity" -> 1

19
tensorflow-ops/tests/GradientTest.hs
View file

@ -230,6 +230,23 @@ testAddNGradient = testCase "testAddNGradient" $ do
TF.gradients y [x] >>= TF.run TF.gradients y [x] >>= TF.run
V.fromList [2, 2, 2 :: Float] @=? dx V.fromList [2, 2, 2 :: Float] @=? dx
testMeanGradient :: Test
testMeanGradient = testCase "testMeanGradient" $ do
[dx] <- TF.runSession $ do
x <- TF.render $ TF.vector [1, 2, 0 :: Float]
let y = TF.mean x (TF.vector [0 :: Int32])
TF.gradients y [x] >>= TF.run
V.fromList [1, 1, 1 :: Float] @=? dx
testMeanGradGrad :: Test
testMeanGradGrad = testCase "testMeanGradGrad" $ do
[ddx] <- TF.runSession $ do
x <- TF.render $ TF.vector [1, 2, 0 :: Float]
let y = TF.mean x (TF.vector [0 :: Int32])
[dx] <- TF.gradients y [x]
TF.gradients dx [x] >>= TF.run
V.fromList [0, 0, 0 :: Float] @=? ddx
testMaxGradient :: Test testMaxGradient :: Test
testMaxGradient = testCase "testMaxGradient" $ do testMaxGradient = testCase "testMaxGradient" $ do
@ -611,6 +628,8 @@ main = defaultMain
, testCreateGraphNameScopes , testCreateGraphNameScopes
, testDiamond , testDiamond
, testAddNGradient , testAddNGradient
, testMeanGradient
, testMeanGradGrad
, testMaxGradient , testMaxGradient
, testConcatGradient , testConcatGradient
, testConcatGradientSimple , testConcatGradientSimple