update_run_pretrain_auto.sh

llm/auto_parallel/deepseek-v3/run_pretrain_auto.sh

@@ -16,32 +16,12 @@
 set -x
 unset CUDA_VISIBLE_DEVICES
 
-mpi_rank=${OMPI_COMM_WORLD_RANK:-0}
-node_rank=$((mpi_rank+offset))
-mpi_node=${OMPI_COMM_WORLD_SIZE:-1}
-echo "MPI status:${mpi_rank}/${mpi_node}"
-nnode_train=${nnode_set:-${mpi_node}}
-master_train=${master:-localhost}
-#
-echo "Distributed Training ${node_rank}/${nnode_train} master=${master_train}"
-set -x
-
-# 屏蔽平台预设的环境变量，因为框架采用兼容升级，检测到这些配置会使用原方式启动
-unset PADDLE_ELASTIC_JOB_ID
-unset PADDLE_TRAINER_ENDPOINTS
-unset DISTRIBUTED_TRAINER_ENDPOINTS
-unset FLAGS_START_PORT
-unset PADDLE_ELASTIC_TIMEOUT
-nnodes=$PADDLE_TRAINERS_NUM
-rank=$PADDLE_TRAINER_ID
-
 task_name="deepseekv3"
 rm -rf output/$task_name/
 rm -rf "output/$task_name""_log"
 
 export SOT_LOG_LEVEL=4
 export PYTHONPATH=../../../:$PYTHONPATH
-export PYTHONPATH=/root/paddlejob/workspace/env_run/xuxinyi/Paddle/build/python:$PYTHONPATH
 #ulimit -c unlimited
 # export GLOG_v=3
 
@@ -95,8 +75,8 @@ python -u  -m paddle.distributed.launch \
     --data_impl "mmap" \
     --enable_auto_parallel 1 \
     --max_grad_norm 1.0 \
-    --num_hidden_layers 5 \
+    --num_hidden_layers 6 \
     --use_intermediate_api true \
     --to_static $to_static \
-    --first_k_dense_replace 6 \
+    --first_k_dense_replace 7 \
     --hidden_size 1792 \