【DPO相关】请问DPO训练结束后，应该如何进行推理呢？能使用类似SFT训练后推理的脚本吗

[Maydaytyh]

Reminder

• I have read the README and searched the existing issues.

System Info

• llamafactory version: 0.9.2.dev0
• Platform: Linux-5.15.0-105-generic-x86_64-with-glibc2.31
• Python version: 3.11.11
• PyTorch version: 2.5.1+cu124 (GPU)
• Transformers version: 4.46.1
• Datasets version: 3.1.0
• Accelerate version: 1.0.1
• PEFT version: 0.12.0
• TRL version: 0.9.6
• GPU type: NVIDIA RTX A6000

Reproduction

这是训练的推理文件

### model
model_name_or_path: mergd_llama2_7b
# trust_remote_code: true

### method
stage: dpo
do_train: true
finetuning_type: lora
lora_target: q_proj,v_proj
pref_beta: 0.1
pref_loss: sigmoid  # choices: [sigmoid (dpo), orpo, simpo]

### dataset
dataset: dataset
template: llama2
# cutoff_len: 2048
# max_samples: 1000
overwrite_cache: true
preprocessing_num_workers: 16

### output
output_dir: new_path
logging_steps: 10
save_steps: 500
plot_loss: true
overwrite_output_dir: true

### train
per_device_train_batch_size: 4
gradient_accumulation_steps: 4
learning_rate: 5.0e-5
num_train_epochs: 1.0
lr_scheduler_type: cosine
warmup_ratio: 0.1
bf16: true
ddp_timeout: 180000000

# ### eval
# val_size: 0.1
# per_device_eval_batch_size: 1
# eval_strategy: steps
# eval_steps: 500(chatkbqa)

这是推理的配置文件

### model
model_name_or_path: mergd_llama2_7b
adapter_name_or_path: new_path

### method
stage: dpo
do_predict: true
finetuning_type: lora
lora_target: q_proj,v_proj

### dataset
template: llama2
eval_dataset: eval_dataset
overwrite_cache: true
preprocessing_num_workers: 16

### output
output_dir: output_dir
logging_steps: 10
save_steps: 1000
plot_loss: true
overwrite_output_dir: true

这样子运行后，无法得到输出文件，所以想请教一下DPO之后应该如何进行推理呢？

Others

No response

[hiyouga]

https://github.com/hiyouga/LLaMA-Factory/tree/main/examples#batch-generation-using-vllm-tensor-parallel