add runtime inferShape in sequence_pool op (PaddlePaddle#41)

zmxdream · Sep 5, 2022 · 5845fd0 · 5845fd0
1 parent 3e8d850
commit 5845fd0
Showing 1 changed file with 44 additions and 3 deletions.
diff --git a/paddle/fluid/operators/sequence_ops/sequence_pool_op.cc b/paddle/fluid/operators/sequence_ops/sequence_pool_op.cc
@@ -27,7 +27,43 @@ class SequencePoolOp : public framework::OperatorWithKernel {
     OP_INOUT_CHECK(ctx->HasInput("X"), "Input", "X", "SequencePool");
     OP_INOUT_CHECK(ctx->HasOutput("Out"), "Output", "Out", "SequencePool");
 
-    if (!ctx->IsRuntime()) {
+    auto ins_dim = ctx->GetInputDim("X");
+    framework::DDim out_dims;
+    PADDLE_ENFORCE_EQ(ins_dim.size(), 2,
+                      platform::errors::InvalidArgument(
+                      "The dims size of first input should be equal to 2, "
+                      "but received value is %d.",
+                      ins_dim.size()));
+
+    if (ctx->IsRuntime()) {
+      auto inputs_tensor = ctx->GetInputVarPtrs("X");
+
+      int batch_size = -1;
+      int rank = ins_dim.size();
+      int cur_batch_size = 0;
+      framework::Variable* x_var =
+          BOOST_GET(framework::Variable*, inputs_tensor[0]);
+        const auto& x_tensor = x_var->Get<LoDTensor>();
+        const auto& x_lod = x_tensor.lod();
+        if (x_lod.size() > 0) {
+          cur_batch_size = x_lod[0].size() - 1;
+        } else {
+          cur_batch_size = x_tensor.dims()[0];
+        }
+        if (batch_size == -1) {
+          batch_size = cur_batch_size;
+        } else {
+          PADDLE_ENFORCE_EQ(batch_size, cur_batch_size,
+                            platform::errors::PreconditionNotMet(
+                                "The batch size of all input should be same, "
+                                "please check, last batch_size is %d, current "
+                                "batch_size is %d",
+                                batch_size, cur_batch_size));
+        }
+        std::vector<int64_t> out_dim;
+        out_dim = {batch_size, ins_dim[rank - 1]};
+        out_dims = phi::make_ddim(out_dim);
+    } else {
       // Check the lod_level for compile-time.
       auto in_lod_level = ctx->GetLoDLevel("X");
       PADDLE_ENFORCE_GT(in_lod_level, 0, platform::errors::InvalidArgument(
@@ -36,13 +72,18 @@ class SequencePoolOp : public framework::OperatorWithKernel {
                                              "lod level %u.",
                                              in_lod_level));
       ctx->SetLoDLevel("Out", in_lod_level - 1);
+
+      int rank = ins_dim.size();
+      std::vector<int64_t> out_dim;
+      out_dim = {-1, ins_dim[rank - 1]};
+      out_dims = phi::make_ddim(out_dim);
     }
 
-    ctx->SetOutputDim("Out", ctx->GetInputDim("X"));
+    ctx->SetOutputDim("Out", out_dims);
     if (ctx->Attrs().Get<std::string>("pooltype") == "MAX") {
       OP_INOUT_CHECK(ctx->HasOutput("MaxIndex"), "Output", "MaxIndex",
                      "SequencePool");
-      ctx->SetOutputDim("MaxIndex", ctx->GetInputDim("X"));
+      ctx->SetOutputDim("MaxIndex", out_dims);
     }
   }
 };