ymcui / Chinese-LLaMA-Alpaca-2

中文LLaMA-2 & Alpaca-2大模型二期项目 + 64K超长上下文模型 (Chinese LLaMA-2 & Alpaca-2 LLMs with 64K long context models)
Apache License 2.0
7.09k stars 578 forks source link

训练完成打印日志的训练时间与实际的训练时间不一致 #314

Closed jiejie1993 closed 1 year ago

jiejie1993 commented 1 year ago

提交前必须检查以下项目

问题类型

模型训练与精调

基础模型

Chinese-LLaMA-2 (7B/13B)

操作系统

Linux

详细描述问题

启动命令:NCCL_IB_DISABLE=1 NCCL_IBEXT_DISABLE=1 NCCL_SOCKET_IFNAME=en,eth,em,bond NCCL_DEBUG=INFO NODE_RANK=0 bash run_pt_ddp_upgrade.sh

run_pt_ddp_upgrade.sh脚本

torchrun --nnodes ${nnodes} --nproc-per-node ${nproc_per_node} --node-rank ${node_rank} --master-addr ${master_address} --master-port ${master_port} \
    --max-restarts ${NUM_ALLOWED_FAILURES} \
    run_clm_pt_with_peft.py \
    --deepspeed ${deepspeed_config_file} \

预训练阶段,使用多机多卡的方式,使用deepspeed-zero2的配置进行训练,硬件是A100,看模型训练完成时的日志是训练了14小时,但实际上消耗了大概约90个小时,请问下,这是因为什么原因?

依赖情况(代码类问题务必提供)

bitsandbytes              0.41.1
peft                      0.3.0.dev0  
sentencepiece             0.1.99
torch                     2.0.1
transformers              4.31.0     

运行日志或截图

{'loss': 2.0443, 'learning_rate': 1.9314285714285718e-05, 'epoch': 0.16}                                                                                                                                   
 16%|█████████████████████████▍                                                                                                                                  | 1142/6990 [8:20:59<121:50:32, 75.01s/it]{'loss': 2.0451, 'learning_rate': 1.9600000000000002e-05, 'epoch': 0.16}                                                                                                                                   
{'loss': 2.0202, 'learning_rate': 1.988571428571429e-05, 'epoch': 0.16}                                                                                                                                    
 16%|█████████████████████████▋                                                                                                                                  | 1151/6990 [8:32:13<122:39:33, 75.62s/it] 16%|█████████████████████████▋                                                                                                                                  | 1153/6990 [8:34:43<122:31:38, 75.57s/i
github-actions[bot] commented 1 year ago

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.

github-actions[bot] commented 1 year ago

Closing the issue, since no updates observed. Feel free to re-open if you need any further assistance.