Dengjia Zhang
Results
1
comments of
Dengjia Zhang
> 有shell不 CUDA_VISIBLE_DEVICES=0 \ swift sft \ --model LLM-Research/Meta-Llama-3.1-8B \ --train_type lora \ --dataset ./DPO_data/2WikimhQA_sft.jsonl \ --torch_dtype bfloat16 \ --num_train_epochs 5 \ --per_device_train_batch_size 1 \ --per_device_eval_batch_size 1 \ --learning_rate 1e-4...