Training prompts

[minji-o-j]

QKV 학습 버전 (tag: QKVo_#num)

• [x] 1번 서버 5 gpu 0, 1

  accelerate launch run_textbox.py --model=PTG --dataset=pc --model_path=facebook/bart-large --gpu_id=0,1 --find_unused_parameters=true --source_task=cross_task1 --training_option=adaptive-attention --QKV_training=True --learning_rate=1e-3

• cross_task1-_05-16-58
• https://wandb.ai/hyu-prompt-transfer/PTG-cross_task-pc/runs/a46s5xiz?workspace=user-minji913

• [x] 5번 서버 5 gpu 2, 3

  accelerate launch run_textbox.py --model=PTG --dataset=pc --model_path=facebook/bart-large --gpu_id=0,1 --find_unused_parameters=true --source_task=cross_dataset2 --training_option=adaptive-attention --QKV_training=True --learning_rate=1e-3

• cross_dataset2--56-03

• https://wandb.ai/hyu-prompt-transfer/PTG-cross_dataset-pc/runs/xrmk4c8v?workspace=user-minji913
  

• [x] 9번 서버 5 gpu 4, 7

  accelerate launch run_textbox.py --model=PTG --dataset=pc --model_path=facebook/bart-large --gpu_id=0,1 --find_unused_parameters=true --source_task=cross_dataset3_paper --training_option=adaptive-attention --QKV_training=True --learning_rate=1e-3

• cross_dataset3_paper--59-25

• https://wandb.ai/hyu-prompt-transfer/PTG-cross_dataset-pc/runs/46o5ky84?workspace=user-minji913
  

• [x] 6번 서버 5 gpu 5, 6

  accelerate launch run_textbox.py --model=PTG --dataset=dd  --model_path=facebook/bart-large --gpu_id=0,1 --find_unused_parameters=true --source_task=cross_dataset2 --training_option=adaptive-attention --QKV_training=True --learning_rate=1e-3

• cross_dataset2--59-17

• https://wandb.ai/hyu-prompt-transfer/PTG-cross_dataset-dd/runs/k889s8ly?workspace=user-minji913
  

• [x] 10번 서버 5 gpu 2, 3

  accelerate launch run_textbox.py --model=PTG --dataset=dd --model_path=facebook/bart-large --gpu_id=0,1 --find_unused_parameters=true --source_task=cross_dataset4_paper --training_option=adaptive-attention --QKV_training=True --learning_rate=1e-3

• cross_dataset4_paper--31-16

• https://wandb.ai/hyu-prompt-transfer/PTG-cross_dataset-dd/runs/ikomvr2c?workspace=user-minji913

[minji-o-j]

QKV 학습X 버전 (tag: QKVx_#num)

• [x] 1번 서버 5 gpu 01

  accelerate launch run_textbox.py --model=PTG --dataset=pc --model_path=facebook/bart-large --gpu_id=0,1 --find_unused_parameters=true --source_task=cross_task1 --training_option=adaptive-attention --QKV_training=False --learning_rate=1e-3

• cross_task1-_14-55-31

• https://wandb.ai/hyu-prompt-transfer/PTG-cross_task-pc/runs/tawkanqv?workspace=user-minji913
  

• [x] 5번 서버 5 gpu 47

  accelerate launch run_textbox.py --model=PTG --dataset=pc --model_path=facebook/bart-large --gpu_id=0,1 --find_unused_parameters=true --source_task=cross_dataset2 --training_option=adaptive-attention --QKV_training=False --learning_rate=1e-3

• cross_dataset2--38-20

• https://wandb.ai/hyu-prompt-transfer/PTG-cross_dataset-pc/runs/k5h651yy?workspace=user-minji913
  

• [x] 9번 서버 5 gpu 23

  accelerate launch run_textbox.py --model=PTG --dataset=pc --model_path=facebook/bart-large --gpu_id=0,1 --find_unused_parameters=true --source_task=cross_dataset3_paper  --training_option=adaptive-attention --QKV_training=False --learning_rate=1e-3

• cross_dataset3_paper--42-47

• https://wandb.ai/hyu-prompt-transfer/PTG-cross_dataset-pc/runs/0100cc55?workspace=user-minji913

• 
  

• [x] 6번 서버 5 gpu 01

  accelerate launch run_textbox.py --model=PTG --dataset=dd  --model_path=facebook/bart-large --gpu_id=0,1 --find_unused_parameters=true --source_task=cross_dataset2 --training_option=adaptive-attention --QKV_training=False --learning_rate=1e-3

• cross_dataset2--17-18

• https://wandb.ai/hyu-prompt-transfer/PTG-cross_dataset-dd/runs/66sl5cwm?workspace=user-minji913
  

• [x] 10번 서버 5 gpu 56

  accelerate launch run_textbox.py --model=PTG --dataset=dd --model_path=facebook/bart-large --gpu_id=0,1 --find_unused_parameters=true --source_task=cross_dataset4_paper --training_option=adaptive-attention --QKV_training=False --learning_rate=1e-3

• cross_dataset4_paper--34-25

• https://wandb.ai/hyu-prompt-transfer/PTG-cross_dataset-dd/runs/up09ykky

[minji-o-j]

Xsum 실험 진행

QKV 학습 버전 (tag: QKVo_#num)

• [x] 4번 서버 5 gpu 0123

• 4개로 학습시켰을때 1에폭당 약 2시간 반 소요

  accelerate launch run_textbox.py --model=PTG --dataset=xsum --model_path=facebook/bart-large --gpu_id=0,1,2,3 --find_unused_parameters=true --source_task=cross_dataset1 --training_option=adaptive-attention --QKV_training=True --learning_rate=1e-3 --train_batch_size=4 --accumulation_steps=24

• cross_dataset1--53-23_QKVo

• test 필요하면 수행해야함

• https://wandb.ai/hyu-prompt-transfer/PTG-cross_dataset-xsum/runs/tfe1x466?workspace=

• [x] 8번 서버 5 gpu 4567

  accelerate launch run_textbox.py --model=PTG --dataset=xsum --model_path=facebook/bart-large --gpu_id=0,1,2,3 --find_unused_parameters=true --source_task=cross_dataset2_paper --training_option=adaptive-attention --QKV_training=True --learning_rate=1e-3 --train_batch_size=4 --accumulation_steps=24

• cross_dataset2_paper--53-29_QKVo

• test 필요하면 수행해야함

• best = 1

• https://wandb.ai/hyu-prompt-transfer/PTG-cross_dataset-xsum/runs/99uu2lps?workspace=user-minji913

QKV 학습x 버전 (tag: QKVx_#num)

• [ ] 4번 서버5 gpu

  accelerate launch run_textbox.py --model=PTG --dataset=xsum --model_path=facebook/bart-large --gpu_id=0,1,2,3 --find_unused_parameters=true --source_task=cross_dataset1 --training_option=adaptive-attention --QKV_training=False --learning_rate=1e-3 --train_batch_size=4 --accumulation_steps=24

• [ ] 8번 서버5 gpu

  accelerate launch run_textbox.py --model=PTG --dataset=xsum --model_path=facebook/bart-large --gpu_id=0,1,2,3 --find_unused_parameters=true --source_task=cross_dataset2_paper --training_option=adaptive-attention --QKV_training=False --learning_rate=1e-3 --train_batch_size=4 --accumulation_steps=24