scripts/enwik8/att_matching.sh

#NUMBER OF HEADS 4
CUDA_VISIBLE_DEVICES=7 python src/lgmodeling/att_matching.py \
    --seed 0 --tgt_len 256 --mem_len 256 --eval_tgt_len 256 \
    --model-a /mnt/data/vinhbk/weights/enwik8/gpt2-finetune/finetune-learnable-indice0-heads4-shared1-routed0-topk0-seed0/best \
    --model-b /mnt/data/vinhbk/weights/enwik8/gpt2-finetune/finetune-learnable-indice0-heads4-shared1-routed0-topk0-seed20/best \
    --data-path /mnt/data/vinhbk/datasets/enwik8 --dataset enwik8
CUDA_VISIBLE_DEVICES=7 python src/lgmodeling/att_matching.py \
    --seed 0 --tgt_len 256 --mem_len 256 --eval_tgt_len 256 \
    --model-a /mnt/data/vinhbk/weights/enwik8/gpt2-finetune/finetune-learnable-indice0-heads4-shared1-routed0-topk0-seed0/best \
    --model-b /mnt/data/vinhbk/weights/enwik8/gpt2-finetune/finetune-learnable-indice0-heads4-shared1-routed0-topk0-seed40/best \
    --data-path /mnt/data/vinhbk/datasets/enwik8 --dataset enwik8

CUDA_VISIBLE_DEVICES=7 python src/lgmodeling/att_matching.py \
    --seed 0 --tgt_len 256 --mem_len 256 --eval_tgt_len 256 \
    --model-a /mnt/data/vinhbk/weights/enwik8/gpt2-finetune/finetune-learnable-indice0-heads4-shared1-routed0-topk0-seed20/best \
    --model-b /mnt/data/vinhbk/weights/enwik8/gpt2-finetune/finetune-learnable-indice0-heads4-shared1-routed0-topk0-seed40/best \
    --data-path /mnt/data/vinhbk/datasets/enwik8 --dataset enwik8

CUDA_VISIBLE_DEVICES=7 python src/lgmodeling/att_matching.py \
    --seed 0 --tgt_len 256 --mem_len 256 --eval_tgt_len 256 \
    --model-a /mnt/data/vinhbk/weights/enwik8/gpt2-finetune/finetune-learnable-indice0-heads8-shared1-routed0-topk0-seed0/best \
    --model-b /mnt/data/vinhbk/weights/enwik8/gpt2-finetune/finetune-learnable-indice0-heads8-shared1-routed0-topk0-seed20/best \
    --data-path /mnt/data/vinhbk/datasets/enwik8 --dataset enwik8
CUDA_VISIBLE_DEVICES=7 python src/lgmodeling/att_matching.py \
    --seed 0 --tgt_len 256 --mem_len 256 --eval_tgt_len 256 \
    --model-a /mnt/data/vinhbk/weights/enwik8/gpt2-finetune/finetune-learnable-indice0-heads8-shared1-routed0-topk0-seed0/best \
    --model-b /mnt/data/vinhbk/weights/enwik8/gpt2-finetune/finetune-learnable-indice0-heads8-shared1-routed0-topk0-seed40/best \
    --data-path /mnt/data/vinhbk/datasets/enwik8 --dataset enwik8

CUDA_VISIBLE_DEVICES=7 python src/lgmodeling/att_matching.py \
    --seed 0 --tgt_len 256 --mem_len 256 --eval_tgt_len 256 \
    --model-a /mnt/data/vinhbk/weights/enwik8/gpt2-finetune/finetune-learnable-indice0-heads8-shared1-routed0-topk0-seed20/best \
    --model-b /mnt/data/vinhbk/weights/enwik8/gpt2-finetune/finetune-learnable-indice0-heads8-shared1-routed0-topk0-seed40/best \
    --data-path /mnt/data/vinhbk/datasets/enwik8 --dataset enwik8



CUDA_VISIBLE_DEVICES=7 python src/lgmodeling/att_matching.py \
    --seed 0 --tgt_len 256 --mem_len 256 --eval_tgt_len 256 \
    --model-a /mnt/data/vinhbk/weights/enwik8/gpt2-finetune/finetune-learnable-indice0-heads16-shared1-routed0-topk0-seed0/best \
    --model-b /mnt/data/vinhbk/weights/enwik8/gpt2-finetune/finetune-learnable-indice0-heads16-shared1-routed0-topk0-seed20/best \
    --data-path /mnt/data/vinhbk/datasets/enwik8 --dataset enwik8
CUDA_VISIBLE_DEVICES=7 python src/lgmodeling/att_matching.py \
    --seed 0 --tgt_len 256 --mem_len 256 --eval_tgt_len 256 \
    --model-a /mnt/data/vinhbk/weights/enwik8/gpt2-finetune/finetune-learnable-indice0-heads16-shared1-routed0-topk0-seed0/best \
    --model-b /mnt/data/vinhbk/weights/enwik8/gpt2-finetune/finetune-learnable-indice0-heads16-shared1-routed0-topk0-seed40/best \
    --data-path /mnt/data/vinhbk/datasets/enwik8 --dataset enwik8

CUDA_VISIBLE_DEVICES=7 python src/lgmodeling/att_matching.py \
    --seed 0 --tgt_len 256 --mem_len 256 --eval_tgt_len 256 \
    --model-a /mnt/data/vinhbk/weights/enwik8/gpt2-finetune/finetune-learnable-indice0-heads16-shared1-routed0-topk0-seed20/best \
    --model-b /mnt/data/vinhbk/weights/enwik8/gpt2-finetune/finetune-learnable-indice0-heads16-shared1-routed0-topk0-seed40/best \
    --data-path /mnt/data/vinhbk/datasets/enwik8 --dataset enwik8