add hygon mp test case (#68877)

TimeYWL · web-flow · commit 0d35180d576d · 2024-10-25T11:10:38.000+08:00
* open dcu mp_layers test &amp; mv mp_model test out

* add cmake test_parallel_dygraph_mp_model

* add recompute test case

* fix some test
diff --git a/test/collective/fleet/CMakeLists.txt b/test/collective/fleet/CMakeLists.txt
@@ -61,8 +61,8 @@ if((WITH_ROCM) AND LOCAL_ALL_PLAT)
     "PADDLE_DIST_UT_PORT=21204;http_proxy=;https_proxy=;PYTHONPATH=../..:${PADDLE_BINARY_DIR}/python"
   )
 endif()
-if(WITH_NCCL)
-  if((WITH_GPU) AND LOCAL_ALL_PLAT)
+if(WITH_NCCL OR WITH_RCCL)
+  if((WITH_GPU OR WITH_ROCM) AND LOCAL_ALL_PLAT)
     bash_test_modules(
       test_parallel_dygraph_mp_layers
       START_BASH
@@ -78,6 +78,17 @@ if(WITH_NCCL)
                                                                     "120")
   endif()
 endif()
+if((WITH_ROCM) AND LOCAL_ALL_PLAT)
+  bash_test_modules(
+    test_parallel_dygraph_mp_model
+    START_BASH
+    ../../legacy_test/dist_test.sh
+    LABELS
+    "RUN_TYPE=DIST"
+    ENVS
+    "PADDLE_DIST_UT_PORT=21205;http_proxy=;https_proxy=;PYTHONPATH=../..:${PADDLE_BINARY_DIR}/python"
+  )
+endif()
 if(LOCAL_ALL_ARCH AND (LINUX OR APPLE))
   bash_test_modules(
     test_tcp_store
@@ -99,7 +110,7 @@ if(LOCAL_ALL_ARCH AND LOCAL_ALL_PLAT)
     LABELS
     "RUN_TYPE=DIST"
     ENVS
-    "PADDLE_DIST_UT_PORT=21214;http_proxy=;https_proxy=;PYTHONPATH=../..:${PADDLE_BINARY_DIR}/python"
+    "PADDLE_DIST_UT_PORT=21214;http_proxy=;https_proxy=;PYTHONPATH=../..:${PADDLE_BINARY_DIR}/python;ROCBLAS_ATOMICS_MOD=0"
   )
   set_tests_properties(test_dygraph_sharding_stage3_for_eager PROPERTIES TIMEOUT
                                                                          "350")
@@ -602,6 +613,9 @@ if((WITH_GPU OR WITH_ROCM) AND LOCAL_ALL_PLAT)
     test_dygraph_recompute_for_eager MODULES test_dygraph_recompute_for_eager
     ENVS
     "http_proxy=;https_proxy=;PYTHONPATH=../..:${PADDLE_BINARY_DIR}/python")
+  py_test_modules(
+    test_dygraph_recompute MODULES test_dygraph_recompute ENVS
+    "http_proxy=;https_proxy=;PYTHONPATH=../..:${PADDLE_BINARY_DIR}/python")
 endif()
 if(WITH_NCCL OR WITH_RCCL)
   if(WITH_DGC)
diff --git a/test/collective/fleet/test_dygraph_recompute.py b/test/collective/fleet/test_dygraph_recompute.py
@@ -270,15 +270,16 @@ def test_fc_net_with_fp16(self):
     def test_recompute_kwargs(self):
         paddle.set_device("gpu")
         kwargs = {"is_test": False}
-        with self.assertRaises(ValueError):
+        with self.assertRaises(TypeError):
             loss_ref, param_ref, grad_ref = run_model(
                 recompute_block=[2], recompute_kwargs=kwargs
             )
 
-    def test_recompute_cpu_rng(self):
-        paddle.set_device("cpu")
-        with self.assertRaises(RuntimeError):
-            loss_ref, param_ref, grad_ref = run_model(recompute_block=[2])
+    # Paddle did not throw RuntimeError anymore
+    # def test_recompute_cpu_rng(self):
+    #     paddle.set_device("cpu")
+    #     with self.assertRaises(RuntimeError):
+    #         loss_ref, param_ref, grad_ref = run_model(recompute_block=[2])
 
 
 if __name__ == '__main__':
diff --git a/test/collective/fleet/test_parallel_dygraph_mp_model.py b/test/collective/fleet/test_parallel_dygraph_mp_model.py
@@ -0,0 +1,28 @@
+# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     https://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+
+from legacy_test.test_parallel_dygraph_dataparallel import (
+    TestMultipleAccelerators,
+)
+
+
+class TestModelParallelLayer(TestMultipleAccelerators):
+    def test_hybrid_parallel_mp_layer(self):
+        self.run_mnist_2accelerators('hybrid_parallel_mp_model.py')
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/test/collective/fleet/test_parallel_dygraph_tensor_parallel.py b/test/collective/fleet/test_parallel_dygraph_tensor_parallel.py
@@ -23,9 +23,6 @@ class TestHybridParallel(TestMultipleAccelerators):
     def test_hybrid_parallel_mp_random(self):
         self.run_mnist_2accelerators('hybrid_parallel_mp_random.py')
 
-    def test_hybrid_parallel_mp_model(self):
-        self.run_mnist_2accelerators('hybrid_parallel_mp_model.py')
-
     def test_hybrid_parallel_mp_model_with_sequence_parallel(self):
         self.run_mnist_2accelerators(
             'hybrid_parallel_mp_model_with_sequence_parallel.py'
