examples: graph: remove trailing transpose/reshape from sdpa example

align with openvino's definition

examples/graph/gpu_opencl_sdpa.cpp

@@ -84,9 +84,6 @@ void gpu_float_sdpa(data_type dtype, int batch_size, int seq_len, int num_head,
     dims qk_output_shape = {batch_size, num_head, seq_len, seq_len};
     dims scale_shape = {1};
     dims attention_mask_shape = {batch_size, 1, 1, seq_len};
-    dims qkv_transpose_order = {0, 2, 1, 3};
-    dims qkv_transposed_shape = {batch_size, seq_len, num_head, size_per_head};
-    dims qkv_reshaped_shape = {batch_size * seq_len, head_dim};
 
     size_t lt_id = 0;
 
@@ -127,28 +124,12 @@ void gpu_float_sdpa(data_type dtype, int batch_size, int seq_len, int num_head,
     op matmul_v {4, op::kind::MatMul, {softmax_out, value_input},
             {matmul_v_out}, "matmul_v"};
 
-    logical_tensor qkv_transposed_out {
-            lt_id++, dtype, qkv_transposed_shape, layout_type::strided};
-    op transpose {5, op::kind::StaticTranspose, {matmul_v_out},
-            {qkv_transposed_out}, "transpose"};
-    transpose.set_attr<std::vector<int64_t>>(
-            op::attr::order, qkv_transpose_order);
-
-    logical_tensor qkv_reshaped_out {
-            lt_id++, dtype, qkv_reshaped_shape, layout_type::strided};
-    op reshape {6, op::kind::StaticReshape, {qkv_transposed_out},
-            {qkv_reshaped_out}, "reshape"};
-    reshape.set_attr(op::attr::special_zero, false);
-    reshape.set_attr<std::vector<int64_t>>(op::attr::shape, qkv_reshaped_shape);
-
     graph g(ekind);
     g.add_op(matmul_qk);
     g.add_op(scale_div);
     g.add_op(mask_add);
     g.add_op(softmax);
     g.add_op(matmul_v);
-    g.add_op(transpose);
-    g.add_op(reshape);
     g.finalize();
 
     std::vector<partition> partitions = g.get_partitions();
@@ -163,7 +144,7 @@ void gpu_float_sdpa(data_type dtype, int batch_size, int seq_len, int num_head,
     std::vector<tensor> inputs_ts, outputs_ts;
     std::vector<std::shared_ptr<void>> data_buffer;
     std::unordered_map<size_t, tensor> global_outputs_ts_map;
-    // Input/output memory should be prepared by users. This helper funciton is
+    // Input/output memory should be prepared by users. This helper function is
     // for testing purpose and not part of API.
     allocate_ocl_graph_mem(
             inputs_ts, inputs, data_buffer, global_outputs_ts_map, eng, true);