File size: 1,657 Bytes
a560a5f |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 |
export LLAVA_DEFAULT_CONVERSATION="conv_vicuna_v0"
# run
export LOGDIR=logs/
export out_dir="${LOGDIR}/llava-plus/llava_plus_v0_7b"
mkdir -p $out_dir
echo ${out_dir}/loginfo.txt
# Note: Our scripts support multi-source data and image folders. Seperate each item with `,`. Note that it may cause problems if multiple folders have images with the same name.
deepspeed llava/train/train_mem.py \
--deepspeed scripts/zero2.json \
--model_name_or_path path/to/LLAVA/vicuna-7b-v0 \
--pretrain_mm_mlp_adapter path/to/LLaVA-7b-pretrain-projector-v0-CC3M-595K-original_caption.bin \
--version v0 \
--data_path path/to/llava-150k-tool-aug.json,path/to/llava-plus-v1-117k-tool-merge.json \
--image_folder /path/to/coco/train2017/,/path/to/hiertext/train,/path/to/infoseek/infoseek_images,/path/to/instruct-pix2pix/clip-filtered-dataset,/path/to/goldg/vg_mdetr/images \
--vision_tower openai/clip-vit-large-patch14 \
--mm_vision_select_layer -2 \
--mm_use_im_start_end False \
--mm_use_im_patch_token False \
--bf16 True \
--output_dir $out_dir \
--num_train_epochs 3 \
--per_device_train_batch_size 16 \
--per_device_eval_batch_size 4 \
--gradient_accumulation_steps 2 \
--evaluation_strategy "no" \
--save_strategy "steps" \
--save_steps 1000 \
--save_total_limit 8 \
--learning_rate 2e-5 \
--weight_decay 0. \
--warmup_ratio 0.03 \
--lr_scheduler_type "cosine" \
--logging_steps 1 \
--tf32 True \
--model_max_length 2048 \
--gradient_checkpointing True \
--dataloader_num_workers 4 \
--lazy_preprocess True \
>> ${out_dir}/loginfo.txt 2>&1 |