pipeline parallel benchmark (#7759)

* pipeline parallel benchmark

* add seed setting

* fixed

tests/test_tipc/dygraph/hybrid_parallelism/ce_gpt/N1C8/CE_gpt-345m_seqlen1024_pretrain_bs32_bf16_PP8-mbs16-acc2.sh

@@ -0,0 +1,43 @@
+# Copyright (c) 2022 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
+#
+#     http://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.
+
+model_item=CE_gpt-345m_seqlen1024_pretrain
+dp_degree=1
+mp_degree=1
+pp_degree=8
+bs_item=32
+fp_item=bf16
+run_mode=PP8-mbs16-acc2
+device_num=N1C8
+max_iter=50000
+
+# default settings of run_benchmark.sh
+use_sharding=False
+sharding_degree=1
+virtual_pp_degree=1
+use_recompute=True
+eval_freq=25
+use_pipeline_parallel=False
+sequence_parallel=False
+
+# For convergence testing
+seed=3589
+model=gpt
+micro_bs=16
+acc_step=2
+
+bash ./test_tipc/dygraph/hybrid_parallelism/ce_gpt/benchmark_common/prepare.sh
+# run
+bash ./test_tipc/dygraph/hybrid_parallelism/ce_gpt/benchmark_common/run_benchmark.sh ${model_item} ${fp_item} ${dp_degree} ${mp_degree} ${pp_degree} ${micro_bs} ${bs_item} ${run_mode} ${device_num} \
+${max_iter} ${use_sharding} ${sharding_degree} ${virtual_pp_degree} ${use_recompute} ${eval_freq} ${use_pipeline_parallel} ${sequence_parallel} ${acc_step} ${seed} 2>&1;

tests/test_tipc/dygraph/hybrid_parallelism/ce_gpt/N1C8/gpt-345m_seqlen1024_pretrain_bs32_bf16_PP8-mbs16-acc2.sh

@@ -0,0 +1,31 @@
+# Copyright (c) 2022 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
+#
+#     http://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.
+
+model_item=gpt-345m_seqlen1024_pretrain
+dp_degree=1
+mp_degree=1
+pp_degree=8
+bs_item=32
+fp_item=bf16
+run_mode=PP8-mbs16-acc2
+device_num=N1C8
+max_iter=100
+
+model=gpt
+micro_bs=16
+
+bash ./test_tipc/dygraph/hybrid_parallelism/ce_gpt/benchmark_common/prepare.sh
+# run
+bash ./test_tipc/dygraph/hybrid_parallelism/ce_gpt/benchmark_common/run_benchmark.sh ${model_item} ${fp_item} ${dp_degree} ${mp_degree} ${pp_degree} ${micro_bs} ${bs_item} ${run_mode} ${device_num} \
+${max_iter} 2>&1;

tests/test_tipc/dygraph/hybrid_parallelism/ce_gpt/benchmark_common/run_benchmark.sh

@@ -153,7 +153,7 @@ function _train(){
             run_pretrain.py ${train_cmd}"
         workerlog_id=0
         ;;
-    DP8-mbs2-acc2|SD8-stage1-mbs2-acc2|SD8-stage2-mbs2-acc2|SD8-stage3-mbs2-acc2|MP2-SD4-stage1-mbs4-acc2|MP2-SP2-PP2-DP2-mbs8-acc2|MP8-mbs16-acc2|MP2-PP2-DP2-mbs8-acc2|MP2-PP2-SD2-Stage1-mbs8-acc2|MP2-SP2-PP2-SD2-Stage1-mbs8-acc2|MP2-SP2-PP4-VPP2-SD2-stage1-mbs2-acc8) echo "run run_mode: ${run_mode}"
+    DP8-mbs2-acc2|SD8-stage1-mbs2-acc2|SD8-stage2-mbs2-acc2|SD8-stage3-mbs2-acc2|PP8-mbs16-acc2|MP2-SD4-stage1-mbs4-acc2|MP2-SP2-PP2-DP2-mbs8-acc2|MP8-mbs16-acc2|MP2-PP2-DP2-mbs8-acc2|MP2-PP2-SD2-Stage1-mbs8-acc2|MP2-SP2-PP2-SD2-Stage1-mbs8-acc2|MP2-SP2-PP4-VPP2-SD2-stage1-mbs2-acc8) echo "run run_mode: ${run_mode}"
         train_cmd="python -m paddle.distributed.launch --log_dir=./mylog --devices=0,1,2,3,4,5,6,7 ${PADDLE_RANK_OPTION}\
             run_pretrain.py ${train_cmd}"
         workerlog_id=0