请问如何指定运行在多个GPU上？ about knowlm HOT 5 CLOSED

(knowlm) me@ubuntugpu:~/KnowLM$ CUDA_VISIBLE_DEVICES=0,1,2,3 python examples/generate_lora.py --base_model zjunlp/knowlm-13b-zhixi --run_ie_cases
testing ie ablities!
load_8bit=False
Loading checkpoint shards: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 3/3 [00:20<00:00,  6.88s/it]
instruction: 我将给你个输入，请根据事件类型列表：['旅游行程']，论元角色列表：['旅游地点', '旅游时间', '旅游人员']，从输入中抽取出可能包含的事件，并以(事件触发词,事件类型,[(事件论元,论元角色)])的形式回答。
input: John昨天在纽约的咖啡馆见到了他的朋友Merry。他们一起喝咖啡聊天，计划着下周去加利福尼亚（California）旅行。他们决定一起租车并预订酒店。他们先计划在下周一去圣弗朗西斯科参观旧金山大桥，下周三去洛杉矶拜访Merry的父亲威廉。
GenerationConfig {
  "num_beams": 4,
  "repetition_penalty": 1.3,
  "temperature": 0.2,
  "top_k": 40,
  "top_p": 0.75,
  "transformers_version": "4.28.1"
}

Traceback (most recent call last):
  File "/mnt/data/me/KnowLM/examples/generate_lora.py", line 217, in <module>
    fire.Fire(main)
  File "/mnt/data/me/.conda/envs/knowlm/lib/python3.9/site-packages/fire/core.py", line 141, in Fire
    component_trace = _Fire(component, args, parsed_flag_args, context, name)
  File "/mnt/data/me/.conda/envs/knowlm/lib/python3.9/site-packages/fire/core.py", line 475, in _Fire
    component, remaining_args = _CallAndUpdateTrace(
  File "/mnt/data/me/.conda/envs/knowlm/lib/python3.9/site-packages/fire/core.py", line 691, in _CallAndUpdateTrace
    component = fn(*varargs, **kwargs)
  File "/mnt/data/me/KnowLM/examples/generate_lora.py", line 212, in main
    print(evaluate(instruction, num_beams=cfg["num_beams"], temperature=cfg["temperature"], repetition_penalty=cfg["repetition_penalty"]))
  File "/mnt/data/me/KnowLM/examples/generate_lora.py", line 193, in evaluate
    generation_output = model.generate(
  File "/mnt/data/me/.conda/envs/knowlm/lib/python3.9/site-packages/torch/autograd/grad_mode.py", line 27, in decorate_context
    return func(*args, **kwargs)
  File "/mnt/data/me/.conda/envs/knowlm/lib/python3.9/site-packages/transformers/generation/utils.py", line 1524, in generate
    return self.beam_search(
  File "/mnt/data/me/.conda/envs/knowlm/lib/python3.9/site-packages/transformers/generation/utils.py", line 2810, in beam_search
    outputs = self(
  File "/mnt/data/me/.conda/envs/knowlm/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1194, in _call_impl
    return forward_call(*input, **kwargs)
  File "/mnt/data/me/.conda/envs/knowlm/lib/python3.9/site-packages/accelerate/hooks.py", line 166, in new_forward
    return module._hf_hook.post_forward(module, output)
  File "/mnt/data/me/.conda/envs/knowlm/lib/python3.9/site-packages/accelerate/hooks.py", line 285, in post_forward
    output = send_to_device(output, self.input_device)
  File "/mnt/data/me/.conda/envs/knowlm/lib/python3.9/site-packages/accelerate/utils/operations.py", line 133, in send_to_device
    return recursively_apply(_send_to_device, tensor, device, non_blocking, test_type=_has_to_method)
  File "/mnt/data/me/.conda/envs/knowlm/lib/python3.9/site-packages/accelerate/utils/operations.py", line 93, in recursively_apply
    {
  File "/mnt/data/me/.conda/envs/knowlm/lib/python3.9/site-packages/accelerate/utils/operations.py", line 94, in <dictcomp>
    k: recursively_apply(
  File "/mnt/data/me/.conda/envs/knowlm/lib/python3.9/site-packages/accelerate/utils/operations.py", line 82, in recursively_apply
    return honor_type(
  File "/mnt/data/me/.conda/envs/knowlm/lib/python3.9/site-packages/accelerate/utils/operations.py", line 53, in honor_type
    return type(obj)(generator)
  File "/mnt/data/me/.conda/envs/knowlm/lib/python3.9/site-packages/accelerate/utils/operations.py", line 85, in <genexpr>
    recursively_apply(
  File "/mnt/data/me/.conda/envs/knowlm/lib/python3.9/site-packages/accelerate/utils/operations.py", line 82, in recursively_apply
    return honor_type(
  File "/mnt/data/me/.conda/envs/knowlm/lib/python3.9/site-packages/accelerate/utils/operations.py", line 53, in honor_type
    return type(obj)(generator)
  File "/mnt/data/me/.conda/envs/knowlm/lib/python3.9/site-packages/accelerate/utils/operations.py", line 85, in <genexpr>
    recursively_apply(
  File "/mnt/data/me/.conda/envs/knowlm/lib/python3.9/site-packages/accelerate/utils/operations.py", line 101, in recursively_apply
    return func(data, *args, **kwargs)
  File "/mnt/data/me/.conda/envs/knowlm/lib/python3.9/site-packages/accelerate/utils/operations.py", line 126, in _send_to_device
    return t.to(device, non_blocking=non_blocking)
torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 18.00 MiB (GPU 0; 10.75 GiB total capacity; 9.45 GiB already allocated; 5.62 MiB free; 9.91 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation.  See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF