Yixin Chen

ValueError: Some specified arguments are not used by the HfArgumentParser: ['--mm_vision_select_layer', '-2', '--mm_use_im_start_end', 'True']

1

I installed the correct transformer library using "pip install --upgrade --force-reinstall git+https://github.com/haotian-liu/transformers_llava.git@26356f0d07bacfb3857dafc7f8a519304b4c0572" but when I run the finetune script, it throws the error message.

some urls in CC-3M Pretrain 595K metadata.json are broken

4

some urls in CC-3M Pretrain 595K metadata.json are broken, could you please provide the downloaded images?

where is the llava_instruct_150k.json

1

Hi, in ''download_data.sh'' we can only download ''complex_reasoning_77k.json conversation_58k.json detail_23k.json'', I wonder if we need to concat them to llava_instruct_150k.json?

assert input_embeddings.shape == embed_tokens_weight.shape

1

Error message: File "/home/ubuntu/project/llava/llava/train/train.py", line 479, in train assert input_embeddings.shape == embed_tokens_weight.shape AssertionError I download the projector weights from https://huggingface.co/liuhaotian/LLaVA-13b-pretrain-projector-v0, and use vicuna-13b as base model.