Gumbel-CRF Semiring (#81)

tests/test_algorithms.py

@@ -3,6 +3,7 @@
     LogSemiring,
     CheckpointSemiring,
     CheckpointShardSemiring,
+    GumbelCRFSemiring,
     KMaxSemiring,
     SparseMaxSemiring,
     MaxSemiring,
@@ -511,3 +512,16 @@ def test_lc_custom():
     # s2 = struct.sum(vals)
     # assert torch.isclose(s, s2).all()
     # assert torch.isclose(marginals, marginals2).all()
+
+
+@given(data())
+def test_gumbel(data):
+    model = data.draw(sampled_from([LinearChain, SemiMarkov, DepTree]))
+    semiring = GumbelCRFSemiring(1.0)
+    test = test_lookup[model]()
+    struct = model(semiring)
+    vals, (batch, N) = test._rand()
+    vals.requires_grad_(True)
+    alpha = struct.marginals(vals)
+    print(alpha[0])
+    print(torch.autograd.grad(alpha, vals, alpha.detach())[0][0])

torch_struct/distributions.py

@@ -16,6 +16,7 @@
     MultiSampledSemiring,
     KMaxSemiring,
     StdSemiring,
+    GumbelCRFSemiring,
 )
 
 
@@ -183,6 +184,13 @@ def count(self):
         ones[self.log_potentials.eq(-float("inf"))] = 0
         return self._struct(StdSemiring).sum(ones, self.lengths)
 
+    def gumbel_crf(self, temperature=1.0):
+        with torch.enable_grad():
+            st_gumbel = self._struct(GumbelCRFSemiring(temperature)).marginals(
+                self.log_potentials, self.lengths
+            )
+            return st_gumbel
+
     # @constraints.dependent_property
     # def support(self):
     #     pass

torch_struct/helpers.py

@@ -104,7 +104,7 @@ def backward(ctx, grad_v):
 
             return DPManual.apply(edge)
 
-    def marginals(self, edge, lengths=None, _autograd=True, _raw=False):
+    def marginals(self, edge, lengths=None, _autograd=True, _raw=False, _combine=False):
         """
         Compute the marginals of a structured model.
 
@@ -135,6 +135,13 @@ def marginals(self, edge, lengths=None, _autograd=True, _raw=False):
                     )
                     all_m.append(self.semiring.unconvert(self._arrange_marginals(marg)))
                 return torch.stack(all_m, dim=0)
+            elif _combine:
+                obj = v.sum(dim=0).sum(dim=0)
+                marg = torch.autograd.grad(
+                    obj, edges, create_graph=True, only_inputs=True, allow_unused=False
+                )
+                a_m = self._arrange_marginals(marg)
+                return a_m
             else:
                 obj = self.semiring.unconvert(v).sum(dim=0)
                 marg = torch.autograd.grad(

torch_struct/semirings/sample.py

@@ -53,6 +53,106 @@ def sum(xs, dim=-1):
         return _SampledLogSumExp.apply(xs, dim)
 
 
+def GumbelMaxSemiring(temp):
+    class _GumbelMaxLogSumExp(torch.autograd.Function):
+        @staticmethod
+        def forward(ctx, input, dim):
+            ctx.save_for_backward(input, torch.tensor(dim))
+            return torch.logsumexp(input, dim=dim)
+
+        @staticmethod
+        def backward(ctx, grad_output):
+            logits, dim = ctx.saved_tensors
+            grad_input = None
+            if ctx.needs_input_grad[0]:
+
+                def sample(ls):
+                    pre_shape = ls.shape
+                    update = (
+                        ls + torch.distributions.Gumbel(0, 1).sample((ls.shape[-1],))
+                    ) / temp
+                    out = torch.nn.functional.one_hot(update.max(-1)[1], pre_shape[-1])
+                    return out
+
+                if dim == -1:
+                    s = sample(logits)
+                else:
+                    dim = dim if dim >= 0 else logits.dim() + dim
+                    perm = [i for i in range(logits.dim()) if i != dim] + [dim]
+                    rev_perm = [
+                        a for a, b in sorted(enumerate(perm), key=lambda a: a[1])
+                    ]
+                    s = sample(logits.permute(perm)).permute(rev_perm)
+
+                grad_input = grad_output.unsqueeze(dim).mul(s)
+            return grad_input, None
+
+    class _GumbelMaxSemiring(_BaseLog):
+        @staticmethod
+        def sum(xs, dim=-1):
+            return _GumbelMaxLogSumExp.apply(xs, dim)
+
+    return _GumbelMaxSemiring
+
+
+def GumbelCRFSemiring(temp):
+    class ST(torch.autograd.Function):
+        @staticmethod
+        def forward(ctx, logits, dim):
+            out = torch.nn.functional.one_hot(logits.max(-1)[1], dim)
+            out = out.type_as(logits)
+            ctx.save_for_backward(logits, out)
+            return out
+
+        @staticmethod
+        def backward(ctx, grad_output):
+            logits, out = ctx.saved_tensors
+            with torch.enable_grad():
+                ret = torch.autograd.grad(
+                    logits.softmax(-1), logits, out * grad_output
+                )[0]
+            return ret, None
+
+    class _GumbelCRFLogSumExp(torch.autograd.Function):
+        @staticmethod
+        def forward(ctx, input, dim):
+            ctx.save_for_backward(input, torch.tensor(dim))
+            return torch.logsumexp(input, dim=dim)
+
+        @staticmethod
+        def backward(ctx, grad_output):
+            logits, dim = ctx.saved_tensors
+            grad_input = None
+            if ctx.needs_input_grad[0]:
+
+                def sample(ls):
+                    update = (
+                        ls + torch.distributions.Gumbel(0, 1).sample((ls.shape[-1],))
+                    ) / temp
+                    out = ST.apply(update, ls.shape[-1])
+                    return out
+
+                if dim == -1:
+                    s = sample(logits)
+                else:
+                    dim = dim if dim >= 0 else logits.dim() + dim
+                    perm = [i for i in range(logits.dim()) if i != dim] + [dim]
+                    rev_perm = [
+                        a for a, b in sorted(enumerate(perm), key=lambda a: a[1])
+                    ]
+                    s = sample(logits.permute(perm)).permute(rev_perm)
+
+                grad_input = grad_output.unsqueeze(dim).mul(s)
+            return grad_input, None
+
+    class _GumbelCRFSemiring(_BaseLog):
+        @staticmethod
+        def sum(xs, dim=-1):
+            return _GumbelCRFLogSumExp.apply(xs, dim)
+
+    return _GumbelCRFSemiring
+
+
 bits = torch.tensor([pow(2, i) for i in range(1, 18)])
 
 

torch_struct/semirings/semirings.py

@@ -277,9 +277,12 @@ class KLDivergenceSemiring(Semiring):
 
     Based on descriptions in:
 
-    * Parameter estimation for probabilistic finite-state transducers :cite:`eisner2002parameter`
-    * First-and second-order expectation semirings with applications to minimum-risk training on translation forests :cite:`li2009first`
+    * Parameter estimation for probabilistic finite-state
+      transducers :cite:`eisner2002parameter`
+    * First-and second-order expectation semirings with applications to
+      minimumrisk training on translation forests :cite:`li2009first`
     * Sample Selection for Statistical Grammar Induction :cite:`hwa2000samplesf`
+
     """
 
     zero = 0