fix(subgraph): fix hand-written backward for serval jit-elemwise ops

GitOrigin-RevId: ea3a40d96e

imperative/python/megengine/functional/nn.py

@@ -909,9 +909,9 @@ def _get_prelu_op(dtype=None, device=None):
         min_0 = f("min", inp, c(0))
         oup = f("fma3", min_0, weight, max_0)
         (oup_grad,) = yield (oup,)
-        inp_grad_0 = f("cond_leq_mov", inp, c(0), oup_grad)
+        inp_grad_0 = f("cond_leq_mov", c(0), inp, oup_grad)
         inp_grad_1 = f("*", oup_grad, weight)
-        inp_grad_1 = f("cond_leq_mov", c(0), inp, inp_grad_1)
+        inp_grad_1 = f("cond_leq_mov", inp, c(0), inp_grad_1)
         inp_grad = f("+", inp_grad_0, inp_grad_1)
         weight_grad = f("*", oup_grad, min_0)
         yield (inp_grad, weight_grad)
@@ -925,7 +925,7 @@ def prelu(inp: Tensor, weight: Tensor) -> Tensor:
     Refer to :class:`~.PReLU` for more information.
     """
     prelu = _get_prelu_op(dtype=inp.dtype, device=inp.device)
-    (oup,) = prelu(inp, weight)
+    (oup,) = prelu(inp, broadcast_to(weight, inp.shape))
     return oup
 
 
@@ -947,7 +947,7 @@ def _get_leagk_relu_op(negative_slope, *, dtype=None, device=None):
         (oup_grad,) = yield (oup,)
         inp_grad_0 = f("cond_leq_mov", c(0), inp, oup_grad)
         inp_grad_1 = f("*", oup_grad, c(negative_slope))
-        inp_grad_1 = f("cond_leq_mov", inp, c(negative_slope), inp_grad_1)
+        inp_grad_1 = f("cond_leq_mov", inp, c(0), inp_grad_1)
         inp_grad = f("+", inp_grad_0, inp_grad_1)
         yield (inp_grad,)
 
@@ -994,13 +994,14 @@ def _get_softplus_op(dtype=None, device=None):
         (inp,) = inputs[0:1]
         neg_abs = f("-", f("abs", inp))
         exp = f("exp", neg_abs)
-        oup = f("log1p", exp)
-        oup = f("+", oup, f("relu", inp))
+        oup0 = f("log1p", exp)
+        oup1 = f("relu", inp)
+        oup = f("+", oup0, oup1)
         (oup_grad,) = yield (oup,)
-        inp_grad_0 = f("switch_gt0", inp, oup_grad)
+        inp_grad_0 = f("switch_gt0", oup1, oup_grad)
         inp_grad_1 = oup_grad
         inp_grad_1 = f("/", oup_grad, f("+", exp, c(1)))
-        inp_grad_1 = f("*", oup_grad, exp)
+        inp_grad_1 = f("*", inp_grad_1, exp)
         inp_grad_1 = f("-", inp_grad_1)
         inp_grad_1 = f("abs_grad", inp, inp_grad_1)
         inp_grad = f("+", inp_grad_0, inp_grad_1)
@@ -1098,16 +1099,17 @@ def _get_logsigmoid_op(dtype=None, device=None):
         (inp,) = inputs[0:1]
         neg_abs = f("-", f("abs", inp))
         exp = f("exp", neg_abs)
-        oup = f("log1p", exp)
-        oup = f("+", oup, f("relu", f("-", inp)))
+        oup0 = f("log1p", exp)
+        oup1 = f("relu", f("-", inp))
+        oup = f("+", oup0, oup1)
         oup = f("-", oup)
         (oup_grad,) = yield (oup,)
         oup_grad = f("-", oup_grad)
-        inp_grad_0 = f("switch_gt0", inp, oup_grad)
+        inp_grad_0 = f("switch_gt0", oup1, oup_grad)
         inp_grad_0 = f("-", inp_grad_0)
         inp_grad_1 = oup_grad
-        inp_grad_1 = f("/", oup_grad, f("+", exp, c(1)))
-        inp_grad_1 = f("*", oup_grad, exp)
+        inp_grad_1 = f("/", inp_grad_1, f("+", exp, c(1)))
+        inp_grad_1 = f("*", inp_grad_1, exp)
         inp_grad_1 = f("-", inp_grad_1)
         inp_grad_1 = f("abs_grad", inp, inp_grad_1)
         inp_grad = f("+", inp_grad_0, inp_grad_1)

imperative/src/impl/ops/utility.cpp

@@ -726,6 +726,28 @@ auto apply_on_var_node(const OpDef& def, const VarNodeArray& inputs) {
 
 auto infer_output_attrs_fallible(
         const OpDef& def, const SmallVector<LogicalTensorDesc>& input_descs) {
+    TensorShape shape;
+    DType dtype = input_descs[0].layout.dtype;
+    CompNode comp_node = input_descs[0].comp_node;
+    for (auto&& desc : input_descs) {
+        if (desc.layout.ndim) {
+            shape = desc.layout;
+            break;
+        }
+    }
+    for (size_t i = 0; i < input_descs.size(); ++i) {
+        if (input_descs[i].layout.ndim) {
+            mgb_assert(
+                    input_descs[i].layout.eq_shape(shape),
+                    "inputs of JITFusionOp should have same shapes");
+        }
+        mgb_assert(
+                input_descs[i].layout.dtype == dtype,
+                "inputs of JITFusionOp should have same dtypes");
+        mgb_assert(
+                input_descs[i].comp_node == comp_node,
+                "inputs of JITFusionOp should have same devices");
+    }
     return OpDef::infer_output_attrs_fallible(
             *def.cast_final_safe<JITFusionOp>().op, input_descs);
 }