Add llama2-70b for test_tipc (#8455)

tests/test_tipc/static/auto_parallel/llama2/N4C32/meta-llama-Llama-2-70b_pretrain_dy2st_bs32_bf16_DP1_MP4_PP8_VPP5.sh

@@ -0,0 +1,26 @@
+# Copyright (c) 2024 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.
+
+param="model_item=meta-llama-Llama-2-70b_pretrain_dy2st "
+param+="run_mode=DP1_MP4_PP8_VPP5 "
+param+="device_num=N4C32 "
+param+="global_batch_size=32 "
+param+="nnodes=4 "
+param+="model_type=llama2_70b "
+
+cd ./tests
+bash ./test_tipc/static/auto_parallel/llama2/benchmark_common/prepare.sh
+
+bash -c "${param} bash ./test_tipc/static/auto_parallel/llama2/benchmark_common/run_benchmark.sh"
+

tests/test_tipc/static/auto_parallel/llama2/pretrain_config_llama2_70b/pretrain-llama2_70b.json

@@ -0,0 +1,62 @@
+{
+    "model_name_or_path": "meta-llama/Llama-2-70b",
+    "tokenizer_name_or_path": "meta-llama/Llama-2-70b",
+    "input_dir": "./data",
+    "output_dir": "./checkpoints/llama2_pretrain_ckpts",
+    "weight_decay": 0.01,
+    "warmup_ratio": 0.01,
+    "max_grad_norm": 1.0,
+    "learning_rate": 3e-05,
+    "min_learning_rate": 3e-06,
+    "warmup_steps": 30,
+    "logging_steps": 1,
+    "max_steps": 50,
+    "save_steps": 5000,
+    "eval_steps": 1000,
+    "continue_training": 0,
+    "do_train": true,
+    "do_eval": false,
+    "do_predict": false,
+    "disable_tqdm": true,
+    "skip_profile_timer": true,
+    "save_total_limit": 2,
+    "device": "gpu",
+    "dataloader_num_workers": 1,
+    "distributed_dataloader": 0,
+    "enable_auto_parallel": true,
+    "per_device_train_batch_size": 1,
+    "gradient_accumulation_steps": 32,
+    "per_device_eval_batch_size": 32,
+    "recompute": false,
+    "recompute_use_reentrant": true,
+    "recompute_granularity": "full",
+    "pp_recompute_interval": 0,
+    "bf16": true,
+    "fp16_opt_level": "O2",
+    "amp_master_grad": true,
+    "amp_custom_black_list": ["reduce_sum", "c_softmax_with_cross_entropy"],
+    "amp_custom_white_list": ["lookup_table", "lookup_table_v2"],
+    "fuse_attention_ffn": true,
+    "fuse_attention_qkv": true,
+    "use_fused_rope": true,
+    "fused_linear_param_grad_add": true,
+    "fuse_sequence_parallel_allreduce": false,
+    "use_flash_attention": true,
+    "use_fused_rms_norm": true,
+    "sep_parallel_degree": 1,
+    "sequence_parallel": true,   
+    "pipeline_parallel_degree": 8,
+    "sharding_parallel_degree": 1,
+    "sharding": "stage1",
+    "tensor_parallel_degree": 4,
+    "virtual_pp_degree": 5,
+    "pipeline_schedule_mode": "VPP", 
+    "data_parallel_config": "enable_allreduce_avg_in_gradinent_scale gradient_sync_after_accumulate",
+    "sharding_parallel_config": "split_param enable_stage1_overlap",
+    "tensor_parallel_config": "enable_mp_async_allreduce",
+    "max_seq_length": 4096,
+    "to_static": true,
+    "eliminate_transpose": 1,
+    "fuse_allreduce_split_to_reducescatter": 1,
+    "sequence_parallel_config": "enable_allreduce_avg_in_gradinent_scale"
+}