Spaces:
Sleeping
Sleeping
File size: 1,164 Bytes
b27b0a2 |
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 |
#!/bin/bash
deepspeed llava/train/train_mem.py \
--deepspeed ./scripts/zero2.json \
--model_name_or_path lmsys/vicuna-13b-v1.5 \
--version plain \
--data_path ./playground/data/LLaVA-Pretrain/blip_laion_cc_sbu_558k.json \
--image_folder ./playground/data/LLaVA-Pretrain/images \
--vision_tower openai/clip-vit-large-patch14-336 \
--mm_projector_type mlp2x_gelu \
--tune_mm_mlp_adapter True \
--mm_vision_select_layer -2 \
--mm_use_im_start_end False \
--mm_use_im_patch_token False \
--bf16 True \
--output_dir ./checkpoints/llava-v1.5-13b-pretrain \
--num_train_epochs 1 \
--per_device_train_batch_size 32 \
--per_device_eval_batch_size 4 \
--gradient_accumulation_steps 1 \
--evaluation_strategy "no" \
--save_strategy "steps" \
--save_steps 24000 \
--save_total_limit 1 \
--learning_rate 1e-3 \
--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 \
--report_to wandb
|