[BUG]binascii.Error: Incorrect padding

Describe the bug
A clear and concise description of what the bug is.
When I running the code deepspeed --num_nodes 1 --num_gpus 4 run_glue_classifier_bert_large.py --task_name RTE --do_train --do_eval --deepspeed --preln --deepspeed_config pwd/glue_bert_large.json --do_lower_case --data_dir /workspace/glue_data/RTE --bert_model bert-large-cased --max_seq_length 128 --train_batch_size 4 --gradient_accumulation_steps 1 --learning_rate 2e-5 --num_train_epochs 3 --output_dir ./output/bert_RTE --model_file 0
To Reproduce
Steps to reproduce the behavior:

1. Go to '...'
2. Click on '....'
3. Scroll down to '....'
4. See error
   Traceback (most recent call last):
   File "/workspace/DeepSpeedExamples/training/BingBertGlue/run_glue_classifier_bert_large.py", line 1260, in
   main()
   File "/workspace/DeepSpeedExamples/training/BingBertGlue/run_glue_classifier_bert_large.py", line 961, in main
   model, optimizer, _, _ = deepspeed.initialize(
   File "/opt/conda/lib/python3.10/site-packages/deepspeed/init.py", line 157, in initialize
   config_class = DeepSpeedConfig(config, mpu)
   File "/opt/conda/lib/python3.10/site-packages/deepspeed/runtime/config.py", line 706, in init
   config_decoded = base64.urlsafe_b64decode(config).decode('utf-8')
   File "/opt/conda/lib/python3.10/base64.py", line 133, in urlsafe_b64decode
   return b64decode(s)
   File "/opt/conda/lib/python3.10/base64.py", line 87, in b64decode
   return binascii.a2b_base64(s)
   Expected behavior
   A clear and concise description of what you expected to happen.

ds_report output
Please run ds_report to give us details about your setup.
[2024-04-08 16:59:46,458] [INFO] [launch.py:315:sigkill_handler] Killing subprocess 7077
[2024-04-08 16:59:46,460] [INFO] [launch.py:315:sigkill_handler] Killing subprocess 7078
[2024-04-08 16:59:46,461] [INFO] [launch.py:315:sigkill_handler] Killing subprocess 7079
[2024-04-08 16:59:46,461] [INFO] [launch.py:315:sigkill_handler] Killing subprocess 7080
[2024-04-08 16:59:46,463] [ERROR] [launch.py:321:sigkill_handler] ['/opt/conda/bin/python', '
Screenshots
If applicable, add screenshots to help explain your problem.

System info (please complete the following information):

• OS: [Ubuntu 20.04]
• GPU count and types [e.g. one machines with 4*NVIDIA RTX A5000 ]
• CUDA Version: 11.7
• Pytorch version 2.0.1
• Any other relevant info about your setup
  Alex

Launcher context
Are you launching your experiment with the deepspeed launcher, MPI, or something else?

Docker context
Are you using a specific docker image that you can share?
2.0.1-cuda11.7-cudnn8
Additional context
Add any other context about the problem here.

### Tasks

Start with deepspeed/runtime/config.py at DeepSpeedConfig and the invocation in run_glue_classifier_bert_large.py, especially the --deepspeed_config argument. Reproduce the command with the supplied environment details and collect ds_report; done means the RTE training command loads its configuration without the binascii padding error.