diff --git a/lib/nnc/ccv_nnc_tensor_io.c b/lib/nnc/ccv_nnc_tensor_io.c
index fa98b9b87..710f01bc6 100644
--- a/lib/nnc/ccv_nnc_tensor_io.c
+++ b/lib/nnc/ccv_nnc_tensor_io.c
@@ -207,8 +207,11 @@ int ccv_nnc_tensor_read(void* const handle, const char* const name, const char*
 							copying = workspace + data_size;
 						else
 							ccv_half_precision_to_float((uint16_t*)(workspace + data_size), (float*)workspace, ccv_min(tensor_count, ccv_min(source_data_size, decoded_size) / sizeof(uint16_t)));
-					} else
+					} else {
+						if (!tensor)
+							*tensor_out = tensor = ccv_nnc_tensor_new(0, tensor_params, 0);
 						ccv_half_precision_to_float((uint16_t*)data, (float*)workspace, ccv_min(tensor_count, sqlite3_column_bytes(tensor_select_stmt, 0) / sizeof(uint16_t)));
+					}
 				} else if (datatype == CCV_32F && tensor_params.datatype == CCV_16F) {
 					size_t decoded_size = source_data_size;
 					if (options->decode(data, sqlite3_column_bytes(tensor_select_stmt, 0), datatype, dim, nd, identifier, options->context, tensor_params, tensor_out, workspace + data_size, &decoded_size))
@@ -217,8 +220,11 @@ int ccv_nnc_tensor_read(void* const handle, const char* const name, const char*
 							copying = workspace + data_size;
 						else
 							ccv_float_to_half_precision((float*)(workspace + data_size), (uint16_t*)workspace, ccv_min(tensor_count, ccv_min(source_data_size, decoded_size) / sizeof(float)));
-					} else
+					} else {
+						if (!tensor)
+							*tensor_out = tensor = ccv_nnc_tensor_new(0, tensor_params, 0);
 						ccv_float_to_half_precision((float*)data, (uint16_t*)workspace, ccv_min(tensor_count, sqlite3_column_bytes(tensor_select_stmt, 0) / sizeof(float)));
+					}
 				} else
 					{ assert(0); }
 			}
@@ -283,7 +289,7 @@ int ccv_nnc_tensor_read(void* const handle, const char* const name, const char*
 				else
 					{ assert(0); }
 			} else {
-				workspace = ccmalloc(data_size + source_data_size);
+				copying = workspace = ccmalloc(data_size + source_data_size);
 				if (datatype == CCV_16F && tensor_params.datatype == CCV_32F)
 				{
 					size_t decoded_size = source_data_size;
@@ -293,8 +299,11 @@ int ccv_nnc_tensor_read(void* const handle, const char* const name, const char*
 							copying = workspace + data_size;
 						else
 							ccv_half_precision_to_float((uint16_t*)(workspace + data_size), (float*)workspace, ccv_min(tensor_count, ccv_min(source_data_size, decoded_size) / sizeof(uint16_t)));
-					} else
+					} else {
+						if (!tensor)
+							*tensor_out = tensor = ccv_nnc_tensor_new(0, tensor_params, 0);
 						ccv_half_precision_to_float((uint16_t*)data, (float*)workspace, ccv_min(tensor_count, sqlite3_column_bytes(tensor_select_stmt, 0) / sizeof(uint16_t)));
+					}
 				} else if (datatype == CCV_32F && tensor_params.datatype == CCV_16F) {
 					size_t decoded_size = source_data_size;
 					if (options->decode(data, sqlite3_column_bytes(tensor_select_stmt, 0), datatype, dim, nd, identifier, options->context, tensor_params, tensor_out, workspace + data_size, &decoded_size))
@@ -303,8 +312,11 @@ int ccv_nnc_tensor_read(void* const handle, const char* const name, const char*
 							copying = workspace + data_size;
 						else
 							ccv_float_to_half_precision((float*)(workspace + data_size), (uint16_t*)workspace, ccv_min(tensor_count, ccv_min(source_data_size, decoded_size) / sizeof(float)));
-					} else
+					} else {
+						if (!tensor)
+							*tensor_out = tensor = ccv_nnc_tensor_new(0, tensor_params, 0);
 						ccv_float_to_half_precision((float*)data, (uint16_t*)workspace, ccv_min(tensor_count, sqlite3_column_bytes(tensor_select_stmt, 0) / sizeof(float)));
+					}
 				} else
 					{ assert(0); }
 			}