rm cmake npu (#53869)

* rm cmake npu

* Update generic.cmake

* Update generic.cmake

cmake/generic.cmake

@@ -575,17 +575,7 @@ function(cc_test_old TARGET_NAME)
     cc_test_build(${TARGET_NAME} SRCS ${cc_test_SRCS} DEPS ${cc_test_DEPS})
     # we dont test hcom op, because it need complex configuration
     # with more than one machine
-    if(NOT
-       ("${TARGET_NAME}" STREQUAL "c_broadcast_op_npu_test"
-        OR "${TARGET_NAME}" STREQUAL "c_allreduce_sum_op_npu_test"
-        OR "${TARGET_NAME}" STREQUAL "c_allreduce_max_op_npu_test"
-        OR "${TARGET_NAME}" STREQUAL "c_reducescatter_op_npu_test"
-        OR "${TARGET_NAME}" STREQUAL "c_allgather_op_npu_test"
-        OR "${TARGET_NAME}" STREQUAL "send_v2_op_npu_test"
-        OR "${TARGET_NAME}" STREQUAL "c_reduce_sum_op_npu_test"
-        OR "${TARGET_NAME}" STREQUAL "recv_v2_op_npu_test"))
-      cc_test_run(${TARGET_NAME} COMMAND ${TARGET_NAME} ARGS ${cc_test_ARGS})
-    endif()
+    cc_test_run(${TARGET_NAME} COMMAND ${TARGET_NAME} ARGS ${cc_test_ARGS})
   elseif(WITH_TESTING AND NOT TEST ${TARGET_NAME})
     add_test(NAME ${TARGET_NAME} COMMAND ${CMAKE_COMMAND} -E echo CI skip
                                          ${TARGET_NAME}.)

python/paddle/distributed/auto_parallel/cluster.py

@@ -29,7 +29,6 @@ class DeviceType(IntEnum):
     CPU = 1
     GPU = 2
     XPU = 3
-    NPU = 4
     DCU = 5
     NIC = 6