Gradients are null when deriving by Variables separately converted to Outputs

[Spiess]

When using Gradients.gradients resulting gradients are null when the Variables are converted to Outputs separately for ys and xs.

Example:

val aVar = tf.variable[Float]("a", Shape(20, 3))
val bVar = tf.variable[Float]("b", Shape(20, 3))

val a: Output[Float] = aVar
val b: Output[Float] = bVar

val xs: Seq[Output[Float]] = Seq(a, b)
val ys: Seq[Output[Float]] = Seq(a - b)

val gradientsAB = Gradients.gradients(ys, xs, Float)

val c = tf.variable[Float]("c", Shape(20, 3))
val d = tf.variable[Float]("d", Shape(20, 3))

val nxs: Seq[Output[Float]] = Seq(c, d)
val nys: Seq[Output[Float]] = Seq(c - d)

val gradientsCD = Gradients.gradients(nys, nxs, Float)

println(gradientsAB)
println(gradientsCD)

Here println(gradientsAB) outputs proper gradients List(Output[...], Output[...]) while println(gradientsCD) outputs List(null, null).

[eaplatanios]

@Spiess This is intentional. In the first case, you create tensors a and b and then define ys as Seq(a - b) and take it's gradients with respect to a and b. In the second case, you define nxs as Seq(c.value, d.value) (explicitly showing the implicit conversion) and nys as Seq(c.value - d.value).c.value call creates a symbolic op in the graph and returns its output tensor. This tensor is different in val nxs: Seq[Output[Float]] = Seq(c, d) and in val nys: Seq[Output[Float]] = Seq(c - d) because two different ops are created. Thus, the nys in this case do not depend on the nxs and their gradients are null. I believe the implicit conversion from variable to tensor is making it hard to notice this distinction.

[Spiess]

Ah right, thanks for clarifying. I assumed that variable value ops still depend on their original variables in the graph, but if that's not the case this output makes sense.

[eaplatanios]

The value ops do depend on the underlying variable but when you take the gradient with respect to the output of the value of, it will only be non-null for tensors that depend on the output of that value op instead of the underlying variable. It is a bit confusing, but I did it this way so that it’s also consistent with the Python API.

I’ll close this issue, but please reopen if a problem persists.