[KVStore] Bug fix (#1480)

* update

* update

* update

* update

* update

* update

* update

* update

* update

* update

* update

* update

* update

* update

* update

* update

* update

* update

* update

* update

* update

* update

* fix lint

* update

* Update network.cc
Co-authored-by: Da Zheng <zhengda1936@gmail.com>

python/dgl/contrib/dis_kvstore.py

@@ -820,7 +820,7 @@ class KVClient(object):
         ----------
         name : str
             data name
-        shape : list of int
+        shape : list or tuple of int
             data shape
         dtype : dtype
             data type
@@ -840,7 +840,7 @@ class KVClient(object):
                 m_id = machines[idx]
                 data_str = self._serialize_shared_tensor(name, dtype)
                 data_str = data_str + '|' + target_name
-                partitioned_shape = shape.copy()
+                partitioned_shape = list(shape)
                 partitioned_shape[0] = count[idx]
                 for n in range(self._group_count):
                     server_id = m_id * self._group_count + n
@@ -1149,6 +1149,7 @@ class KVClient(object):
 
         We usually invoke this API by just one client (e.g., client_0).
         """
+        if self._client_id == 0:
             for server_id in range(self._server_count):
                 msg = KVStoreMsg(
                     type=KVMsgType.FINAL,

src/graph/network.cc

@@ -65,6 +65,8 @@ NDArray CreateNDArrayFromRaw(std::vector<int64_t> shape,
 }
 
 void ArrayMeta::AddArray(const NDArray& array) {
+  // Get data type of current NDArray
+  data_type_.push_back(array->dtype);
   // We first write the ndim to the data_shape_
   data_shape_.push_back(static_cast<int64_t>(array->ndim));
   // Then we write the data shape
@@ -82,6 +84,9 @@ char* ArrayMeta::Serialize(int64_t* size) {
     buffer_size += sizeof(ndarray_count_);
     buffer_size += sizeof(data_shape_.size());
     buffer_size += sizeof(int64_t) * data_shape_.size();
+    // we don't need to write data_type_.size()
+    // because it equals to ndarray_count_ * 3
+    buffer_size += sizeof(DLDataType) * data_type_.size();
   }
   // In the future, we should have a better memory management as
   // allocating a large chunk of memory can be very expensive.
@@ -94,6 +99,11 @@ char* ArrayMeta::Serialize(int64_t* size) {
     // Write ndarray_count_
     *(reinterpret_cast<int*>(pointer)) = ndarray_count_;
     pointer += sizeof(ndarray_count_);
+    // Write data type
+    memcpy(pointer,
+        reinterpret_cast<DLDataType*>(data_type_.data()),
+        sizeof(DLDataType) * data_type_.size());
+    pointer += (sizeof(DLDataType) * data_type_.size());
     // Write size of data_shape_
     *(reinterpret_cast<size_t*>(pointer)) = data_shape_.size();
     pointer += sizeof(data_shape_.size());
@@ -117,6 +127,12 @@ void ArrayMeta::Deserialize(char* buffer, int64_t size) {
     ndarray_count_ = *(reinterpret_cast<int*>(buffer));
     buffer += sizeof(int);
     data_size += sizeof(int);
+    // Read data type
+    data_type_.resize(ndarray_count_);
+    memcpy(data_type_.data(), buffer,
+        ndarray_count_ * sizeof(DLDataType));
+    buffer += ndarray_count_ * sizeof(DLDataType);
+    data_size += ndarray_count_ * sizeof(DLDataType);
     // Read size of data_shape_
     size_t count = *(reinterpret_cast<size_t*>(buffer));
     buffer += sizeof(size_t);
@@ -570,7 +586,7 @@ static KVStoreMsg* recv_kv_message(network::Receiver* receiver) {
     CHECK_EQ(meta.data_shape_[0], 1);
     kv_msg->id = CreateNDArrayFromRaw(
       {meta.data_shape_[1]},
-      DLDataType{kDLInt, 64, 1},
+      meta.data_type_[0],
       DLContext{kDLCPU, 0},
       recv_id_msg.data,
       AUTO_FREE);
@@ -588,7 +604,7 @@ static KVStoreMsg* recv_kv_message(network::Receiver* receiver) {
     }
     kv_msg->data = CreateNDArrayFromRaw(
       vec_shape,
-      DLDataType{kDLFloat, 32, 1},
+      meta.data_type_[1],
       DLContext{kDLCPU, 0},
       recv_data_msg.data,
       AUTO_FREE);
@@ -607,7 +623,7 @@ static KVStoreMsg* recv_kv_message(network::Receiver* receiver) {
     }
     kv_msg->shape = CreateNDArrayFromRaw(
       vec_shape,
-      DLDataType{kDLInt, 64, 1},
+      meta.data_type_[0],
       DLContext{kDLCPU, 0},
       recv_shape_msg.data,
       AUTO_FREE);
@@ -772,7 +788,7 @@ DGL_REGISTER_GLOBAL("network._CAPI_FastPull")
         kv_msg.rank = client_id;
         kv_msg.name = name;
         kv_msg.id = CreateNDArrayFromRaw({static_cast<int64_t>(remote_ids[i].size())},
-                                         DLDataType{kDLInt, 64, 1},
+                                         ID->dtype,
                                          DLContext{kDLCPU, 0},
                                          remote_ids[i].data(),
                                          !AUTO_FREE);
@@ -812,7 +828,7 @@ DGL_REGISTER_GLOBAL("network._CAPI_FastPull")
     local_data_shape[0] = ID_size;
     NDArray res_tensor = CreateNDArrayFromRaw(
                           local_data_shape,
-                          DLDataType{kDLFloat, 32, 1},
+                          local_data->dtype,
                           DLContext{kDLCPU, 0},
                           return_data,
                           AUTO_FREE);

src/graph/network.h

@@ -67,6 +67,7 @@ enum MessageType {
   kIPIDMsg = 7
 };
 
+
 /*!
  * \brief Meta data for NDArray message
  */
@@ -133,6 +134,11 @@ class ArrayMeta {
    */
   int ndarray_count_;
 
+  /*!
+   * \brief DataType for each NDArray
+   */
+  std::vector<DLDataType> data_type_;
+
   /*!
    * \brief We first write the ndim to data_shape_ 
    * and then write the data shape. 

tests/compute/test_kvstore.py

@@ -23,15 +23,27 @@ data_1 = F.zeros((num_entries*2, dim_size), F.float32, F.cpu())
 g2l_1 = F.arange(0, num_entries*2)
 partition_1 = F.zeros(num_entries*2, F.int64, F.cpu())
 
+data_3 = F.zeros((num_entries, dim_size), F.int64, F.cpu())
+data_4 = F.zeros((num_entries, dim_size), F.float64, F.cpu())
+data_5 = F.zeros((num_entries, dim_size), F.int32, F.cpu())
+
 def start_server():
     my_server = KVServer(server_id=0, server_namebook=server_namebook, num_client=1)
-
     my_server.set_global2local(name='data_0', global2local=g2l_0)
     my_server.set_global2local(name='data_1', global2local=g2l_1)
+    my_server.set_global2local(name='data_3', global2local=g2l_0)
+    my_server.set_global2local(name='data_4', global2local=g2l_0)
+    my_server.set_global2local(name='data_5', global2local=g2l_0)
     my_server.set_partition_book(name='data_0', partition_book=partition_0)
     my_server.set_partition_book(name='data_1', partition_book=partition_1)
+    my_server.set_partition_book(name='data_3', partition_book=partition_0)
+    my_server.set_partition_book(name='data_4', partition_book=partition_0)
+    my_server.set_partition_book(name='data_5', partition_book=partition_0)
     my_server.init_data(name='data_0', data_tensor=data_0)
     my_server.init_data(name='data_1', data_tensor=data_1)
+    my_server.init_data(name='data_3', data_tensor=data_3)
+    my_server.init_data(name='data_4', data_tensor=data_4)
+    my_server.init_data(name='data_5', data_tensor=data_5)
 
     my_server.start()
 
@@ -40,13 +52,17 @@ def start_client():
     my_client = KVClient(server_namebook=server_namebook)
     my_client.connect()
 
-    my_client.init_data(name='data_2', shape=[num_entries, dim_size], dtype=F.float32, target_name='data_0')
+    my_client.init_data(name='data_2', shape=(num_entries, dim_size), dtype=F.float32, target_name='data_0')
+    print("Init data from client..")
 
     name_list = my_client.get_data_name_list()
-    assert len(name_list) == 3
+    assert len(name_list) == 6
     assert 'data_0' in name_list
     assert 'data_1' in name_list
     assert 'data_2' in name_list
+    assert 'data_3' in name_list
+    assert 'data_4' in name_list
+    assert 'data_5' in name_list
 
     meta_0 = my_client.get_data_meta('data_0')
     assert meta_0[0] == F.float32
@@ -60,8 +76,23 @@ def start_client():
     assert meta_2[0] == F.float32
     assert_array_equal(meta_2[2], partition_0)
 
+    meta_3 = my_client.get_data_meta('data_3')
+    assert meta_3[0] == F.int64
+    assert_array_equal(meta_3[2], partition_0)    
+
+    meta_4 = my_client.get_data_meta('data_4')
+    assert meta_4[0] == F.float64
+    assert_array_equal(meta_3[2], partition_0)  
+
+    meta_5 = my_client.get_data_meta('data_5')
+    assert meta_5[0] == F.int32
+    assert_array_equal(meta_3[2], partition_0)  
+
     my_client.push(name='data_0', id_tensor=F.tensor([0, 1, 2]), data_tensor=F.tensor([[1.,1.,1.],[2.,2.,2.],[3.,3.,3.]]))
     my_client.push(name='data_2', id_tensor=F.tensor([0, 1, 2]), data_tensor=F.tensor([[1.,1.,1.],[2.,2.,2.],[3.,3.,3.]]))
+    my_client.push(name='data_3', id_tensor=F.tensor([0, 1, 2]), data_tensor=F.tensor([[1,1,1],[2,2,2],[3,3,3]]))
+    my_client.push(name='data_4', id_tensor=F.tensor([0, 1, 2]), data_tensor=F.tensor([[1.,1.,1.],[2.,2.,2.],[3.,3.,3.]], F.float64))
+    my_client.push(name='data_5', id_tensor=F.tensor([0, 1, 2]), data_tensor=F.tensor([[1,1,1],[2,2,2],[3,3,3]], F.int32))
 
     target = F.tensor([[1.,1.,1.],[2.,2.,2.],[3.,3.,3.]])
 
@@ -71,6 +102,21 @@ def start_client():
     res = my_client.pull(name='data_2', id_tensor=F.tensor([0, 1, 2]))
     assert_array_equal(res, target)
 
+    target = F.tensor([[1,1,1],[2,2,2],[3,3,3]])
+
+    res = my_client.pull(name='data_3', id_tensor=F.tensor([0, 1, 2]))
+    assert_array_equal(res, target)
+
+    target = F.tensor([[1.,1.,1.],[2.,2.,2.],[3.,3.,3.]], F.float64)
+
+    res = my_client.pull(name='data_4', id_tensor=F.tensor([0, 1, 2]))
+    assert_array_equal(res, target)
+
+    target = F.tensor([[1,1,1],[2,2,2],[3,3,3]], F.int32)
+
+    res = my_client.pull(name='data_5', id_tensor=F.tensor([0, 1, 2]))
+    assert_array_equal(res, target)
+
     my_client.shut_down()