deepspeed zero2/zero3 训练报错：KeyError: 'use_muon'

[Jintao-Huang]

感谢moonlight如此优秀模型的开源，swift借助examples/toy_train.py的实现集成了muon优化器：modelscope/ms-swift#3234

shell可以参考这里：https://github.com/modelscope/ms-swift/blob/main/examples/train/optimizer/muon.sh

但是在进行deepspeed zero2/zero3训练时，遇到报错：KeyError: 'use_muon'，后续会兼容zero2/zero3的训练嘛

复现脚本：

NPROC_PER_NODE=2 \
CUDA_VISIBLE_DEVICES=0,1 \
swift sft \
    --model Qwen/Qwen2.5-7B-Instruct \
    --train_type lora \
    --dataset 'AI-ModelScope/alpaca-gpt4-data-zh#500' \
              'AI-ModelScope/alpaca-gpt4-data-en#500' \
              'swift/self-cognition#500' \
    --optimizer muon \
    --torch_dtype bfloat16 \
    --num_train_epochs 1 \
    --per_device_train_batch_size 1 \
    --per_device_eval_batch_size 1 \
    --learning_rate 1e-4 \
    --lora_rank 8 \
    --lora_alpha 32 \
    --target_modules all-linear \
    --gradient_accumulation_steps 16 \
    --eval_steps 50 \
    --save_steps 50 \
    --save_total_limit 5 \
    --logging_steps 5 \
    --max_length 2048 \
    --output_dir output \
    --system 'You are a helpful assistant.' \
    --warmup_ratio 0.05 \
    --dataloader_num_workers 4 \
    --model_author swift \
    --model_name swift-robot \
    --deepspeed zero2

[toothacher17]

Hi, thanks for trying to migrate it to Deepspeed. However, maintaining a code in Deepseed is not in our roadmap, but we do plan to support for Megatron-LM

Please see an example PR here: NVIDIA/Megatron-LM#1428