[Usage] Not able to import llava file while fine tunning

[Vish2427]

Describe the issue

Issue: While running the fine tune script its not able to import the llava file when running the train_mem.py file. Command:

deepspeed llava/train/train_mem.py \
    --lora_enable True --lora_r 128 --lora_alpha 256 --mm_projector_lr 2e-5 \
    --bits 4 \
    --deepspeed {DEEPSPEED_JSON} \
    --model_name_or_path {MODEL_NAME} \
    --version v1 \
    --data_path {DATA_PATH} \
    --image_folder {IMAGE_FOLDER} \
    --vision_tower {VISION_TOWER} \
    --mm_projector_type mlp2x_gelu \
    --mm_vision_select_layer -2 \
    --mm_use_im_start_end False \
    --mm_use_im_patch_token False \
    --image_aspect_ratio pad \
    --group_by_modality_length True \
    --bf16 True \
    --output_dir {OUTPUT_DIR} \
    --num_train_epochs 5 \
    --per_device_train_batch_size 16 \
    --per_device_eval_batch_size 4 \
    --gradient_accumulation_steps 1 \
    --evaluation_strategy "no" \
    --save_strategy "steps" \
    --save_steps 250 \
    --save_total_limit 1 \
    --learning_rate 2e-4 \
    --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

Screenshots:

[Bobby-youngking]

change the import path, there are so many wrong paths in the code

[moaldeen]

change the import path, there are so many wrong paths in the code i finished finetuning but when i feed it the same image i used for fine tuning it still predict it wrong

[devsangho]

pip install -e .

You can use the command as provided in Readme.