Can I call the forward function in parallel during the inference phase?

I inited a model.
Can I call function model(args) in parallel (e.g. multiple requests) during inference ？
I got the error when qps >= 2

File "/home/web_server/src/processor.py", line 41, in process
res = self.model.predict(input)
File "/home/web_server/predict_en.py", line 379, in predict
res = forward_step_func(self.model, input)
File "/home/web_server/src/predict_en.py", line 89, in forward_step_func
output_tensor = model(input),
File "/home/web_server/anaconda3/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1130, in _call_impl
return forward_call(*input, **kwargs)
File "/home/web_server/anaconda3/lib/python3.8/site-packages/deepspeed/utils/nvtx.py", line 11, in wrapped_fn
return func(*args, **kwargs)
File "/home/web_server/anaconda3/lib/python3.8/site-packages/deepspeed/runtime/engine.py", line 1668, in forward
self.start_timers(self.engine_timers.forward_timers)
File "/home/web_server/anaconda3/lib/python3.8/site-packages/deepspeed/runtime/engine.py", line 2081, in start_timers
self.timers(name).start()
File "/home/web_server/anaconda3/lib/python3.8/site-packages/deepspeed/utils/timer.py", line 44, in start
assert not self.started, f"{self.name} timer has already been started"
AssertionError: forward_microstep timer has already been started

Start with processor.py:41 and predict_en.py at predict and forward_step_func, then inspect DeepSpeed runtime/engine.py:1668 and utils/timer.py:44. Reproduce concurrent inference calls at QPS >= 2 and determine whether shared model state or the forward_microstep timer causes the assertion. Done means the supported concurrency behavior and the cause are established, with a regression test or documented limitation.