what's the difference between BertForSequenceClassification.from_pretrained and python run_glue.py?

I found two fine tune ways for Sequence Classification with transformer:

1,BertForSequenceClassification.from_pretrained(https://huggingface.co/transformers/training.html):

from transformers import BertForSequenceClassification
model = BertForSequenceClassification.from_pretrained('bert-base-uncased')
model.train()

2,python run_glue.py(https://github.com/huggingface/transformers/tree/master/examples/text-classification):

export GLUE_DIR=/path/to/glue
export TASK_NAME=MRPC

python run_glue.py \
  --model_name_or_path bert-base-cased \
  --task_name $TASK_NAME \
  --do_train \
  --do_eval \
  --data_dir $GLUE_DIR/$TASK_NAME \
  --max_seq_length 128 \
  --per_device_train_batch_size 32 \
  --learning_rate 2e-5 \
  --num_train_epochs 3.0 \
  --output_dir /tmp/$TASK_NAME/

Are there any difference between them?