LoRA Model Separate Export During HF Conversion #5204
Description
I am exporting a LoRA model in HF format using MS-SWIFT.
Is there an option to export the base model and the LoRA model separately?
The merged model is too large, so I plan to build the service by swapping only the LoRA.
However, even when setting merge to false during model export, it seems that a separate LoRA file is not generated.
Could you please confirm whether there is currently an option to export them separately? (_ _)
我正在使用 MS-SWIFT 以 HF 格式导出 LoRA 模型。
请问是否有选项可以将基础模型和 LoRA 模型分别导出？
合并后的模型太大，因此我计划通过替换 LoRA 来构建服务。
但是，即使在导出模型时将 merge 设置为 false，也似乎没有生成单独的 LoRA 文件。
请问目前是否支持将它们分别导出的选项？(_ _)
Train Parameters
Memory usage: 8 * 78GiB
Training speed: 9.5s/it
PYTORCH_CUDA_ALLOC_CONF='expandable_segments:True'
NPROC_PER_NODE=8
CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
megatron sft
--load Qwen3-235B-A22B-Instruct-2507-mcore
--dataset 'swift/Chinese-Qwen3-235B-2507-Distill-data-110k-SFT#2000'
'swift/self-cognition#1000'
--optimizer_cpu_offload true
--use_precision_aware_optimizer true
--train_type lora
--lora_rank 8
--lora_alpha 32
--target_modules all-linear
--split_dataset_ratio 0.01
--expert_model_parallel_size 2
--pipeline_model_parallel_size 4
--decoder_first_pipeline_num_layers 23
--decoder_last_pipeline_num_layers 23
--moe_grouped_gemm true
--moe_shared_expert_overlap true
--moe_aux_loss_coeff 1e-3
--micro_batch_size 8
--global_batch_size 16
--recompute_granularity full
--recompute_method uniform
--recompute_num_layers 1
--max_epochs 1
--finetune true
--cross_entropy_loss_fusion true
--lr 1e-4
--lr_warmup_fraction 0.05
--min_lr 1e-5
--save megatron_output/Qwen3-235B-A22B-Instruct-2507
--eval_interval 200
--save_interval 200
--max_length 2048
--num_workers 8
--dataset_num_proc 8
--no_save_optim true
--no_save_rng true
--sequence_parallel true
--attention_backend flash
--model_author swift
--model_name swift-robot