LMFlow icon indicating copy to clipboard operation
LMFlow copied to clipboard

Unable to resolve any data

Open ShiXiangXiang123 opened this issue 2 years ago • 2 comments

deepspeed ${deepspeed_args}
examples/finetune.py
--deepspeed configs/ds_config_zero3.json
--bf16
--run_name finetune_with_lora
--model_name_or_path /output_models/llama7b-lora-170k
--num_train_epochs 0.01
--learning_rate 2e-5
--dataset_path /data/converted_data.json
--per_device_train_batch_size 1
--per_device_eval_batch_size 1
--validation_split_percentage 0
--logging_steps 20
--block_size 512
--do_train
--output_dir output_models/my_first_model
--overwrite_output_dir
--ddp_timeout 72000
--save_steps 5000
--dataloader_num_workers 1

这是我的数据: image 格式是text2text

报错: image

ShiXiangXiang123 avatar Jun 14 '23 12:06 ShiXiangXiang123

这是我的数据格式: image qing'd请大神帮我看看哪里出问题了

ShiXiangXiang123 avatar Jun 14 '23 15:06 ShiXiangXiang123

数据路径指向json文件所在的文件夹路径即可,比如 --dataset_path /data 建议用绝对路径


The data path should point to the directory where the JSON file is located, for example --dataset_path /data. It is recommended to use an absolute path.

shizhediao avatar Jun 19 '23 10:06 shizhediao

This issue has been marked as stale because it has not had recent activity. If you think this still needs to be addressed please feel free to reopen this issue. Thanks

shizhediao avatar Jul 09 '23 14:07 shizhediao