You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This discussion was converted from issue #5863 on October 30, 2024 06:16.
Heading
Bold
Italic
Quote
Code
Link
Numbered list
Unordered list
Task list
Attach files
Mention
Reference
Menu
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
-
Reminder
System Info
llamafactory
version: 0.9.1.dev0Reproduction
llamafactory-cli train --stage pt --do_train True --model_name_or_path /root/autodl-tmp/Qwen/Qwen2___5-1___5B-Instruct --preprocessing_num_workers 16 --finetuning_type full --template qwen --flash_attn auto --dataset_dir data --dataset class_stand --cutoff_len 4096 --learning_rate 1e-05 --num_train_epochs 20.0 --max_samples 10000 --per_device_train_batch_size 2 --gradient_accumulation_steps 8 --lr_scheduler_type linear --max_grad_norm 0.5 --logging_steps 5 --save_steps 200 --warmup_steps 200 --optim adamw_torch --packing True --report_to none --use_galore True --output_dir /root/autodl-tmp/Qwen2.5/full/train_2024-10-29-11-37-39 --bf16 True --plot_loss True --ddp_timeout 180000000 --include_num_input_tokens_seen True --galore_rank 8 --galore_update_interval 200 --galore_scale 0.25 --galore_target all --val_size 0.1 --eval_strategy steps --eval_steps 200 --per_device_eval_batch_size 2 --overwrite_output_dir --save_total_limit 1 --load_best_model_at_end True
Expected behavior
我的问题是:通常大语言模型的知识量是在预训练过程注入的,而预训练通常会将长文本切割成一个个段落。但是这样的话在训练时是怎么关联起来这个上下文的呢?例如:我的数据集中有一个30k长度左右的文本,我想让继续预训练后的模型完全记住我这个文本中的所有内容,我应该如何做呢?
Others
No response
Beta Was this translation helpful? Give feedback.
All reactions