[huggingface/transformers]Transformers训练器没有属性“deepspeed_plugin”

@shibing624我们需要更多信息。你的...是accelerate env？你到底是如何运行这个脚本的？请向我们提供这些关键细节，以便我们重现

加速环境：

- `Accelerate` version: 0.23.0
- Platform: Linux-5.4.119-1-tlinux4-0009.3-x86_64-with-glibc2.17
- Python version: 3.10.11
- Numpy version: 1.24.4
- PyTorch version (GPU?): 2.0.1+cu117 (True)
- PyTorch XPU available: False
- PyTorch NPU available: False
- System RAM: 1006.96 GB
- GPU type: A100-SXM4-40GB
- `Accelerate` default config:
        - compute_environment: LOCAL_MACHINE
        - distributed_type: MULTI_GPU
        - mixed_precision: fp16
        - use_cpu: False
        - debug: False
        - num_processes: 8
        - machine_rank: 0
        - num_machines: 1
        - gpu_ids: all
        - rdzv_backend: static
        - same_network: True
        - main_training_function: main
        - downcast_bf16: no
        - tpu_use_cluster: False
        - tpu_use_sudo: False
        - tpu_env: []

ds_报告：

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]
fused_adam ............. [NO] ....... [OKAY]
cpu_adam ............... [NO] ....... [OKAY]
cpu_adagrad ............ [NO] ....... [OKAY]
cpu_lion ............... [NO] ....... [OKAY]
 [WARNING]  Please specify the CUTLASS repo directory as environment variable $CUTLASS_PATH
evoformer_attn ......... [NO] ....... [NO]
fused_lamb ............. [NO] ....... [OKAY]
fused_lion ............. [NO] ....... [OKAY]
quantizer .............. [NO] ....... [OKAY]
random_ltd ............. [NO] ....... [OKAY]
 [WARNING]  sparse_attn requires a torch version >= 1.5 and < 2.0 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]
--------------------------------------------------
DeepSpeed general environment info:
torch install path ............... ['/apdcephfs_teg_2/share_1367250/flemingxu/miniconda3/envs/py3.10/lib/python3.10/site-packages/torch']
torch version .................... 2.0.1+cu117
deepspeed install path ........... ['/apdcephfs_teg_2/share_1367250/flemingxu/miniconda3/envs/py3.10/lib/python3.10/site-packages/deepspeed']
deepspeed info ................... 0.11.1, unknown, unknown
torch cuda version ............... 11.7
torch hip version ................ None
nvcc version ..................... 11.7
deepspeed wheel compiled w. ...... torch 2.0, cuda 11.7
shared memory (/dev/shm) size .... 503.48 GB

我的训练论据是：

@dataclass
class PeftArguments(TrainingArguments):
    use_peft: bool = field(default=True, metadata={"help": "Whether to use peft"})
    target_modules: Optional[str] = field(default="all")
    lora_rank: Optional[int] = field(default=8)
    lora_dropout: Optional[float] = field(default=0.05)
    lora_alpha: Optional[float] = field(default=32.0)
    modules_to_save: Optional[str] = field(default=None)
    peft_path: Optional[str] = field(default=None, metadata={"help": "The path to the peft model"})
    qlora: bool = field(default=False, metadata={"help": "Whether to use qlora"})
    load_in_kbits: Optional[int] = field(default=None, metadata={"help": "Kbits to train the model, value is 4, 8"})
    model_max_length: int = field(
        default=512,
        metadata={"help": "Maximum sequence length. suggest value is 8192 * 4, 8192 * 2, 8192, 4096, 2048, 1024, 512"}
    )

我正在尝试修复它，我添加了 deepspeed_plugin 并调试到 PeftArguments，错误已修复。

    deepspeed_plugin: Optional[str] = field(default=None)
    debug: Optional[str] = field(
        default="",
        metadata={
            "help": (
                "Whether or not to enable debug mode. default is '', "
                "`underflow_overflow` (Detect underflow and overflow in activations and weights), "
            )
        },
    )

我的仓库修复了错误：https://github.com/shibing624/MedicalGPT/commit/b08be905d7f2b98a3e62c57bb6e8b345c0611805

原因：错误在这里： https: //github.com/huggingface/transformers/blob/main/src/transformers/trainer.py#L1174C30-L1174C30

我不使用 Deepspeed，错误它在这里运行。我不知道如何修复原始变形金刚 trainer.py

真正的解决方案是我们应该None默认将其设置为TrainingArguments，有一种方法可以使其成为隐藏属性，我会研究这个，除非你想 @shibing624