Fix single run error for gpt model

official/nlp/gpt/scripts/run_standalone_train.sh

@@ -30,4 +30,5 @@ python train.py  \
     --epoch_size=$EPOCH_SIZE \
     --device_id=$DEVICE_ID \
     --data_path=$DATA_DIR \
+    --model_parallel=1 \
     --optimizer="adam" > training_log.txt 2>&1 &

official/nlp/pangu_alpha/README.md

@@ -241,7 +241,7 @@ Training 60B model using 8 NPU in one server requires that the server has at lea
 ```bash
 # run distributed training example in one ascend machine
 
-bash run_distributed_train_moe_host_device.sh /path/dataset /path/hccl.json 8 fp32 2.6B 1 1 1 0 8 36 0
+bash run_distributed_train_moe_host_device.sh /path/dataset /path/hccl.json 8 fp32 2.6B 1 1 2 0 8 36 0
 ```
 
 #### Training on homogeneous