Closed smallBy closed 5 months ago
训练指令如下: torchrun --nnodes 1 --nproc_per_node 8 run_clm_sft_with_peft.py \ --deepspeed ds_zero2_no_offload.json \ --model_name_or_path chinese-alpaca-2-13b/ \ --tokenizer_name_or_path scripts/tokenizer/ \ --dataset_dir data/ \ --per_device_train_batch_size 1 \ --per_device_eval_batch_size 1 \ --do_train \ --seed $RANDOM \ --fp16 \ --num_train_epochs 2 \ --lr_scheduler_type cosine \ --learning_rate 1e-4 \ --warmup_ratio 0.03 \ --weight_decay 0 \ --logging_strategy steps \ --logging_steps 10 \ --save_strategy steps \ --save_total_limit 3 \ --evaluation_strategy steps \ --eval_steps 250 \ --save_steps 500 \ --gradient_accumulation_steps 1 \ --preprocessing_num_workers 8 \ --max_seq_length 512 \ --output_dir output/ \ --overwrite_output_dir \ --ddp_timeout 30000 \ --logging_first_step True \ --lora_rank 64 \ --lora_alpha 128 \ --trainable "q_proj,v_proj,k_proj,o_proj,gate_proj,down_proj,up_proj" \ --modules_to_save "embed_tokens,lm_head" \ --lora_dropout 0.05 \ --torch_dtype float16 \ --save_safetensors False \ --validation_file data/validate.json \ --load_in_kbits 16 \ --save_safetensors False \ --gradient_checkpointing \ --ddp_find_unused_parameters False
系统:ubuntu20.04 GPU为P100
This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your consideration.
Closing the issue, since no updates observed. Feel free to re-open if you need any further assistance.
提交前必须检查以下项目
问题类型
None
基础模型
None
操作系统
None
详细描述问题
依赖情况(代码类问题务必提供)
运行日志或截图