Skip to content
This repository has been archived by the owner on Oct 12, 2023. It is now read-only.

2400条数据,10个epoch,pre_seq_len=128,lora训练,为什么推理时显示trainable params: 0 || all params: 6243584000 || trainable%: 0.0000,是数据量不够吗,下边贴出训练参数 #396

Closed
zeze813 opened this issue Aug 9, 2023 · 1 comment
Labels
solved This problem has been already solved.

Comments

@zeze813
Copy link

zeze813 commented Aug 9, 2023

CUDA_VISIBLE_DEVICES=2 python3 src/train_bash.py
--stage sft
--model_name_or_path $path_to_your_chatglm_model
--do_train
--dataset_dir $data_dir
--dataset ww
--finetuning_type lora
--output_dir $path_to_sft_checkpoint
--per_device_train_batch_size 2
--gradient_accumulation_steps 4
--lr_scheduler_type cosine
--logging_steps 10
--save_steps 500
--learning_rate $lr
--num_train_epochs $epoch
--pre_seq_len $pre_seq_len
--lora_rank 32
--fp16

@zeze813 zeze813 changed the title 2400条数据,10个epoch,lora训练,推理时显示trainable params: 0 || all params: 6243584000 || trainable%: 0.0000 2400条数据,10个epoch,lora训练,为什么推理时显示trainable params: 0 || all params: 6243584000 || trainable%: 0.0000,是数据量不够吗,下边贴出训练参数 Aug 9, 2023
@zeze813 zeze813 changed the title 2400条数据,10个epoch,lora训练,为什么推理时显示trainable params: 0 || all params: 6243584000 || trainable%: 0.0000,是数据量不够吗,下边贴出训练参数 2400条数据,10个epoch,pre_seq_len=128,lora训练,为什么推理时显示trainable params: 0 || all params: 6243584000 || trainable%: 0.0000,是数据量不够吗,下边贴出训练参数 Aug 9, 2023
@hiyouga hiyouga added the solved This problem has been already solved. label Aug 9, 2023
@hiyouga
Copy link
Owner

hiyouga commented Aug 9, 2023

推理本身就不应该有可训练参数

@hiyouga hiyouga closed this as completed Aug 9, 2023
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
solved This problem has been already solved.
Projects
None yet
Development

No branches or pull requests

2 participants