[XPU] EmbeddingWithEltwiseAddXpuKernel support FP16 (#51426)

paddle/phi/backends/xpu/xpu1_op_list.cc

@@ -81,7 +81,7 @@ XPUOpMap& get_kl1_ops() {
       {"elementwise_sub_grad", XPUKernelSet({phi::DataType::FLOAT32})},
       {"elementwise_sub", XPUKernelSet({phi::DataType::FLOAT32})},
       {"embedding_with_eltwise_add_xpu",
-       XPUKernelSet({phi::DataType::FLOAT32})},
+       XPUKernelSet({phi::DataType::FLOAT32, phi::DataType::FLOAT16})},
       {"equal", XPUKernelSet({phi::DataType::INT64})},
       {"expand_as_v2",
        XPUKernelSet({phi::DataType::INT32,

paddle/phi/backends/xpu/xpu2_op_list.cc

@@ -225,7 +225,7 @@ XPUOpMap& get_kl2_ops() {
                      phi::DataType::INT64,
                      phi::DataType::INT32})},
       {"embedding_with_eltwise_add_xpu",
-       XPUKernelSet({phi::DataType::FLOAT32})},
+       XPUKernelSet({phi::DataType::FLOAT32, phi::DataType::FLOAT16})},
       {"empty",
        XPUKernelSet({phi::DataType::INT64,
                      phi::DataType::INT32,

paddle/phi/kernels/fusion/xpu/embedding_with_eltwise_add_xpu_kernel.cc

@@ -25,6 +25,7 @@ void EmbeddingWithEltwiseAddXpuKernel(
     const std::vector<const DenseTensor*>& tables,
     int64_t padding_idx,
     DenseTensor* out) {
+  using XPUType = typename XPUTypeTrait<T>::Type;
   auto& id_dims = ids[0]->dims();
   int idx_len = id_dims[0] * id_dims[1];
   int emb_layer_num = ids.size();
@@ -47,8 +48,21 @@ void EmbeddingWithEltwiseAddXpuKernel(
             table_dims[1],
             embed_dim));
     table_lens_cpu.push_back(table_dims[0]);
+    if (std::is_same<T, phi::dtype::float16>::value) {
+      DenseTensor table_data_fp32_t;
+      ctx.template Alloc<float>(&table_data_fp32_t,
+                                table->numel() * sizeof(float));
+      int r = xpu::cast<XPUType, float>(
+          ctx.x_context(),
+          reinterpret_cast<const XPUType*>(table->data<T>()),
+          table_data_fp32_t.data<float>(),
+          table->numel());
+      PADDLE_ENFORCE_XDNN_SUCCESS(r, "cast");
+      arg_tables.push_back(table_data_fp32_t.data<float>());
+    } else {
       arg_tables.push_back(table->data<float>());
     }
+  }
   std::vector<std::vector<int>> int_idx(emb_layer_num,
                                         std::vector<int>(idx_len, 0));
   std::vector<xpu::VectorParam<int>> arg_ids;
@@ -70,17 +84,39 @@ void EmbeddingWithEltwiseAddXpuKernel(
     arg_ids.push_back(
         xpu::VectorParam<int>{int_idx[i].data(), idx_len, nullptr});
   }
+
   ctx.template Alloc<T>(out);
+  if (std::is_same<T, phi::dtype::float16>::value) {
+    DenseTensor out_fp32_t;
+    ctx.template Alloc<float>(&out_fp32_t, out->numel() * sizeof(float));
+    int r = xpu::multi_embedding_fusion<float, float, int>(
+        ctx.x_context(),
+        arg_tables, /* tables */
+        out_fp32_t.data<float>(),
+        arg_ids,
+        table_lens_cpu,
+        embed_dim,
+        std::vector<float>(table_lens_cpu.size(), 1.0f),
+        std::vector<int>(table_lens_cpu.size(), padding_idx));
+    PADDLE_ENFORCE_XDNN_SUCCESS(r, "embedding_with_eltwise_add_xpu");
+
+    r = xpu::cast(ctx.x_context(),
+                  out_fp32_t.data<float>(),
+                  reinterpret_cast<float16*>(out->data<T>()),
+                  out->numel());
+    PADDLE_ENFORCE_XDNN_SUCCESS(r, "cast");
+  } else {
     int r = xpu::multi_embedding_fusion<float, float, int>(
         ctx.x_context(),
         arg_tables, /* tables */
-      out->data<T>(),
+        out->data<float>(),
         arg_ids,
         table_lens_cpu,
         embed_dim,
         std::vector<float>(table_lens_cpu.size(), 1.0f),
         std::vector<int>(table_lens_cpu.size(), padding_idx));
     PADDLE_ENFORCE_XDNN_SUCCESS(r, "embedding_with_eltwise_add_xpu");
+  }
 }
 
 }  // namespace fusion
@@ -90,6 +126,7 @@ PD_REGISTER_KERNEL(embedding_with_eltwise_add_xpu,
                    XPU,
                    ALL_LAYOUT,
                    phi::fusion::EmbeddingWithEltwiseAddXpuKernel,
-                   float) {
+                   float,
+                   phi::dtype::float16) {
   kernel->InputAt(0).SetBackend(phi::Backend::CPU);
 }