run_train.sh

python -m torch.distributed.launch --nproc_per_node=2 --master_port=34235 train.py  /mnt/msranlp/shaohanh/data/tnlg_config/            --task image_gpt_pretraining_wild_laion_v3_obj          --tokens-per-sample 2048          --criterion unigpt          --arch unigptmodel_xl              --required-batch-size-multiple 1              --optimizer adam                --adam-betas '(0.9,0.98)'                --adam-eps 1e-6                --clip-norm 2.0               --lr-scheduler polynomial_decay                --weight-decay 0.01                --lr 0.0001                --warmup-updates 375                --total-num-update 30000               --max-update 30000               --max-sentences 2               --update-freq 2               --log-format simple      --log-interval 50     --disable-validation               --save-interval-updates 1000     --no-epoch-checkpoints               --memory-efficient-fp16     --fp16-init-scale 4     --fp16-scale-window 256               --min-loss-scale 0.0001               --seed 2               --dict-path /mnt/msranlp/shumma/data/16g/dict.txt                --spm-model /mnt/msranlp/shumma/data/16g/sentencepiece.bpe.model               --save-dir  /mnt/msranlpintern/dataset/cornell-instruction/dataloader/08/train_output/01/model_savedir          --tensorboard-logdir  /mnt/msranlpintern/dataset/cornell-instruction-v1/dataloader/08/train_output/01/tensorboard_logdir         --init-from-file /mnt/msranlp/shaohanh/exp/unigpt_exp/vl_xl_oailrg_mtnlg_6e4_256-2048_deepnorm_1M_xcon_wild_laion_coyo_rerun_v2/checkpoint_1_300000.pt          --ddp-backend=no_c10d               --distributed-no-spawn               --batch-read-ahead 100               --reset-dataloader           --train-json-split-name train-nogithub-noarvix-nopubmed-mtnlg          --image-encoder clip   --visual-model-name ViT-L-14 --visual-output-dim  1024          --visual-pretrained /mnt/msranlp/shaohanh/exp/unigpt_exp/data/models/openai_clip/ViT-L-14-sd.pt            --wild-data-dir /mnt/msranlp/shaohanh/bvl/wild_subset1/            --wild-batch-size 1          --laion-data-dir /mnt/msranlpintern/dataset/cornell-instruction-v1/dataloader/08/dataloader_config         --laion-batch-size 12          --quantized-size 32          --locate-special-token 1          --box-score-threshold 0.65          --mix-no-object-prob 0.3          --latent-query-num 64 --connector xconnector          --no-freeze-layer resblocks.23,ln_post          --subln         --sope-rel-pos          --data-weights 0,1,0          --checkpoint-activations