[RELAY][GRAD] handle Tuple/TupleGetItem in first order gradient (apac…

…he#5946) * handle Tuple/TupleGetItem in first order gradient * Unify MultiOnes/MultiZeros.
trevor-m · Jun 30, 2020 · ff23786 · ff23786
1 parent 0189bca
commit ff23786
Show file tree

Hide file tree

Showing 3 changed files with 104 additions and 9 deletions.
diff --git a/src/relay/transforms/gradient.cc b/src/relay/transforms/gradient.cc
@@ -106,14 +106,44 @@ struct ADValueNode {
   }
 };
 
+template <typename F>
+Expr MultiFactory(const Type& t, F factory) {
+  if (auto* tt = t.as<TensorTypeNode>()) {
+    return factory(tt->shape, tt->dtype);
+  } else if (auto* tt = t.as<TupleTypeNode>()) {
+    std::vector<Expr> res;
+    for (size_t i = 0; i < tt->fields.size(); i++) {
+      res.push_back(MultiFactory(tt->fields[i], factory));
+    }
+    return Tuple(res);
+  } else {
+    LOG(FATAL) << "unsupported type to create tensors of: " << tt;
+    throw;
+  }
+}
+
+template <typename F, typename F2>
+Expr MultiFactoryLike(const Expr& e, const Type& t, F factory, F2 factory_like) {
+  if (t.as<TensorTypeNode>()) {
+    return factory_like(e);
+  } else if (auto* tt = t.as<TupleTypeNode>()) {
+    return MultiFactory(t, factory);
+  } else {
+    LOG(FATAL) << "unsupported type to tensors of: " << tt;
+    throw;
+  }
+}
+
 using ADValue = std::shared_ptr<ADValueNode>;
 
 /*! \brief AD over a program which generates a tensor output. */
 struct ADTensor : ADValueNode {
   Expr forward;
   mutable Expr reverse;  // must be a variable to avoid duplication
   ADTensor(LetList* ll, const Expr& forward)
-      : forward(ll->Push(forward)), reverse(ll->Push(ZerosLike(this->forward))) {
+      : forward(ll->Push(forward)),
+        reverse(
+            ll->Push(MultiFactoryLike(this->forward, forward->checked_type(), Zeros, ZerosLike))) {
     this->forward->checked_type_ = forward->checked_type();
   }
 };
@@ -165,6 +195,51 @@ struct FirstOrderReverseAD : ExprFunctor<ADValue(const Expr&)> {
         });
   }
 
+  ADValue VisitExpr_(const TupleGetItemNode* op) final {
+    Expr e = GetRef<Expr>(op);
+    ADValue tup = VisitExpr(op->tuple);
+    auto tt = op->tuple->checked_type().as<TupleTypeNode>();
+    size_t size = tt->fields.size();
+    size_t idx = op->index;
+    auto ret = std::make_shared<ADTensor>(ll, e);
+    backprop_actions.push_back([tup, idx, size, ret](LetList* ll) {
+      auto rev = tup->get<ADTensor>().reverse;
+      // special-case Tuple, to avoid long chains of GetItem/Tuple,
+      // but we might have functions using tuples, so we don't know
+      // that the reverse node is always a tuple
+      std::vector<Expr> grfields;
+      if (auto tup_node = rev.as<TupleNode>()) {
+        for (size_t i = 0; i < size; ++i) {
+          grfields.push_back(i != idx ? tup_node->fields[i]
+                                      : Add(tup_node->fields[i], ret->reverse));
+        }
+      } else {
+        for (size_t i = 0; i < size; ++i) {
+          grfields.push_back(i != idx ? TupleGetItem(rev, i)
+                                      : Add(TupleGetItem(rev, i), ret->reverse));
+        }
+      }
+      tup->get<ADTensor>().reverse = ll->Push(Tuple(grfields));
+    });
+    return ret;
+  }
+
+  ADValue VisitExpr_(const TupleNode* op) final {
+    Expr e = GetRef<Expr>(op);
+    std::vector<ADValue> fields;
+    for (const auto& f : op->fields) {
+      fields.push_back(VisitExpr(f));
+    }
+    auto ret = std::make_shared<ADTensor>(ll, e);
+    backprop_actions.push_back([fields, ret](LetList* ll) {
+      for (size_t i = 0; i < fields.size(); ++i) {
+        fields[i]->get<ADTensor>().reverse =
+            ll->Push(Add(fields[i]->get<ADTensor>().reverse, TupleGetItem(ret->reverse, i)));
+      }
+    });
+    return ret;
+  }
+
   ADValue VisitExpr_(const ConstantNode* op) final {
     Expr e = GetRef<Expr>(op);
     return std::make_shared<ADTensor>(ll, e);
@@ -235,7 +310,7 @@ Expr FirstOrderGradient(const Expr& re, const Optional<IRModule>& mod) {
     auto c = rev->get<ADFunction>().func(f->checked_type(), args, Attrs(), {});
     const auto& res = c->get<ADTensor>();
     Expr grad = LetList::With([&](LetList* ll) {
-      res.reverse = OnesLike(res.forward);
+      res.reverse = MultiFactoryLike(res.forward, res.forward->checked_type(), Ones, OnesLike);
       for (auto it = reverse_ad.backprop_actions.rbegin(); it != reverse_ad.backprop_actions.rend();
            ++it) {
         (*it)(ll);

diff --git a/src/relay/transforms/pattern_util.h b/src/relay/transforms/pattern_util.h
@@ -524,6 +524,12 @@ inline Expr OnesLike(Expr e) {
   return Call(op, {e});
 }
 
+Expr MakeOnes(Expr shape, DataType dtype);
+
+inline Expr Ones(Array<IndexExpr> shape, DataType dtype) {
+  return MakeOnes(CheckConstantShape(shape), dtype);
+}
+
 inline Expr CollapseSumLike(Expr e) {
   static const Op& op = Op::Get("collapse_sum_like");
   return Call(op, {e});

diff --git a/tests/python/relay/test_pass_gradient.py b/tests/python/relay/test_pass_gradient.py
@@ -158,20 +158,27 @@ def test_broadcast_subtract():
                                 -np.ones_like(expected_forward).sum(axis=(0, 1), keepdims=True).squeeze(axis=0))
 
 
-def test_tuple():
+def _test_tuple(mode):
     shape = (10, 10)
     dtype = 'float32'
     t = relay.TensorType(shape, dtype)
     x = relay.var("x", t)
     y = relay.var("y", t)
     z = relay.var("z", t)
-    tup = relay.Var("tup")
-    func = relay.Function([x, y, z], relay.Let(tup, relay.Tuple([x, y, z]),
-                                               relay.TupleGetItem(tup, 0) +
-                                               relay.TupleGetItem(tup, 1) -
-                                               relay.TupleGetItem(tup, 2)))
+    if mode == "higher_order":
+        tup = relay.Var("tup")
+        func = relay.Function([x, y, z], relay.Let(tup, relay.Tuple([x, y, z]),
+                                                   relay.TupleGetItem(tup, 0) +
+                                                   relay.TupleGetItem(tup, 1) -
+                                                   relay.TupleGetItem(tup, 2)))
+    else:
+        # first order does not do let.
+        tup = relay.Tuple([x, y, z])
+        func = relay.Function([x, y, z], relay.TupleGetItem(tup, 0) +
+                                         relay.TupleGetItem(tup, 1) -
+                                         relay.TupleGetItem(tup, 2))
     func = run_infer_type(func)
-    back_func = run_infer_type(gradient(func))
+    back_func = run_infer_type(gradient(func, mode=mode))
     assert back_func.checked_type == relay.FuncType([t, t, t], relay.TupleType([t, relay.TupleType([t, t, t])]))
     x_nd = rand(dtype, *shape)
     y_nd = rand(dtype, *shape)
@@ -188,6 +195,12 @@ def test_tuple():
     tvm.testing.assert_allclose(grad_z.asnumpy(), -1 * np.ones_like(grad_z.asnumpy()))
 
 
+def test_tuple():
+    _test_tuple("higher_order")
+
+def test_tuple_first_order():
+    _test_tuple("first_order")
+
 def test_pow():
     mod = tvm.IRModule()
     p = Prelude(mod)
@@ -304,6 +317,7 @@ def test_concat():
     test_broadcast_add()
     test_broadcast_subtract()
     test_tuple()
+    test_tuple_first_order()
     test_pow()
     test_ref()
     test_square_second_order()