Concurrent generation of responses (one GPU, multiple users)

Describe the bug
Can Deepspeed handle multiple inference calls at the same time? That means multiple simultaneous calls to model.generate(...) on one GPU. Or are multiple concurrent calls to the tokenizer a problem?

To Reproduce

import transformers
from transformers import GPTJForCausalLM
import torch
import deepspeed
tokenizer = transformers.AutoTokenizer.from_pretrained("EleutherAI/gpt-j-6B", pad_token='<|endoftext|>', eos_token='<|endoftext|>', truncation_side='left')
model = GPTJForCausalLM.from_pretrained(
            "EleutherAI/gpt-j-6B"
            revision="float16",
            torch_dtype=torch.float16,
            low_cpu_mem_usage=True,
            use_cache=True,
            gradient_checkpointing=True,
)
model = deepspeed.init_inference(
            model=model,
            mp_size=1,
            dtype=model.dtype,
            replace_method="auto",
            replace_with_kernel_inject=True
)
device = 'cuda'
model.to(device)

prompt = tokenizer("This is a test sentence", return_tensors='pt', truncation="only_first", max_length=2048)
prompt = {key: value.to(device) for key, value in prompt.items()}
out = model.generate(**prompt,
                                  min_length=16,
                                  max_new_tokens=100,
                                  do_sample=True,
                                  top_k=35,
                                  top_p=0.6,
                                  temperature=1,
                                  no_repeat_ngram_size=4,
                                  use_cache=True,
                                  pad_token_id=tokenizer.eos_token_id,
                                  stopping_criteria=StoppingCriteriaList([stopping_criteria])  # Here I provide some stopping criteria
)
res = tokenizer.decode(out[0])
torch.cuda.empty_cache()

Expected behavior
Successful concurrent calls to tokenizer(...)and model.generate(...).

ds_report output

DeepSpeed C++/CUDA extension op report

NOTE: Ops not installed will be just-in-time (JIT) compiled at
runtime if needed. Op compatibility means that your system
meet the required dependencies to JIT install the op.

JIT compiled ops requires ninja
ninja .................. [OKAY]

op name ................ installed .. compatible

[WARNING] async_io requires the dev libaio .so object and headers but these were not found.
[WARNING] async_io: please install the libaio-dev package with apt
[WARNING] If libaio is already installed (perhaps from source), try setting the CFLAGS and LDFLAGS environment variables to where it can be found.
async_io ............... [NO] ....... [NO]
cpu_adagrad ............ [NO] ....... [OKAY]
cpu_adam ............... [NO] ....... [OKAY]
fused_adam ............. [NO] ....... [OKAY]
fused_lamb ............. [NO] ....... [OKAY]
quantizer .............. [NO] ....... [OKAY]
random_ltd ............. [NO] ....... [OKAY]
[WARNING] sparse_attn requires a torch version >= 1.5 but detected 2.0
[WARNING] using untested triton version (2.0.0), only 1.0.0 is known to be compatible
sparse_attn ............ [NO] ....... [NO]
spatial_inference ...... [NO] ....... [OKAY]
transformer ............ [NO] ....... [OKAY]
stochastic_transformer . [NO] ....... [OKAY]
transformer_inference .. [NO] ....... [OKAY]
utils .................. [NO] ....... [OKAY]

DeepSpeed general environment info:
torch install path ............... ['/home/myUsername/anaconda3/envs/huggingface/lib/python3.11/site-packages/torch']
torch version .................... 2.0.0+cu117
deepspeed install path ........... ['/home/myUsername/anaconda3/envs/huggingface/lib/python3.11/site-packages/deepspeed']
deepspeed info ................... 0.8.3, unknown, unknown
torch cuda version ............... 11.7
torch hip version ................ None
nvcc version ..................... 11.7
deepspeed wheel compiled w. ...... torch 2.0, cuda 11.8

System info (please complete the following information):

• OS: Ubuntu 22.04 LTS
• One machine with 1x Nvidia RTX 3090
• Deepspeed 0.8.3
• Transformers 4.27.2
• Accelerate 0.17.1
• Python v3.11

Start with the provided Python reproducer, focusing on concurrent tokenizer(...) and model.generate(...) calls on one CUDA device. Check the behavior with the listed DeepSpeed, Transformers, PyTorch, and CUDA versions; done would require a documented concurrency result or a clearly scoped fix supported by a test.