wisconsinaivision / vip-llava Goto Github PK

View Code? Open in Web Editor NEW

167.0 5.0 8.0 17.78 MB

[CVPR2024] ViP-LLaVA: Making Large Multimodal Models Understand Arbitrary Visual Prompts

Home Page: https://vip-llava.github.io/

License: Apache License 2.0

Dockerfile 0.30% Shell 7.22% Python 88.72% HTML 1.48% JavaScript 1.92% CSS 0.35%

chatbot clip foundation-models gpt-4 gpt-4-vision llama llama2 llava multi-modal vision-language

vip-llava's Introduction

🎯 [CVPR2024] ViP-LLaVA: Making Large Multimodal Models Understand Arbitrary Visual Prompts

Mu Cai, Haotian Liu, Siva Karthik Mustikovela, Gregory P. Meyer, Yuning Chai, Dennis Park, Yong Jae Lee

Approach: [Project Page] [Demo] [Model Zoo] [Paper]

ViP-Bench: [Download Dataset] [LeaderBoard] [Evaluation Server]

Release

[04/26] 🔥 LLaVA and ViP-LLaVA with the recent Llama-3-8B and Phi-3-mini-3.8B LLM backbones is available here!
[02/26] 🔥 ViP-LLaVA is accepted to CVPR 2024!
[12/13] 🔥 Our works now appears on the official Huggingface transformers doc!
[12/03] 🔥 We released ViP-LLaVA: Making Large Multimodal Models Understand Arbitrary Visual Prompts. We propose to directly overlay the visual prompts upon the the original image during visual instruction tunning, so that large multimodal models could possibly understand arbitrary visual prompts in a user-friendly way. Checkout the paper and demo. We also built the first zero-shot region-level benchmark ViP-Bench for large multimodal models.

Usage and License Notices: The data and checkpoint is intended and licensed for research use only. They are also restricted to uses that follow the license agreement of LLaMA, Vicuna and GPT-4. The dataset is CC BY NC 4.0 (allowing only non-commercial use) and models trained using the dataset should not be used outside of research purposes.

Install
ViP-LLaVA Weights
Demo
Model Zoo
Dataset
Train
Evaluation

Install

If you are not using Linux, do NOT proceed, see instructions for macOS and Windows.

Clone this repository and navigate to ViP-LLaVA folder

cd ViP-LLaVA

Install Package

conda create -n vip-llava python=3.10 -y
conda activate vip-llava
pip install --upgrade pip  # enable PEP 660 support
pip install -e .

Install additional packages for training cases

pip install -e ".[train]"
pip install flash-attn --no-build-isolation

Quick Start With HuggingFace

Example Code

from llava.model.builder import load_pretrained_model
from llava.mm_utils import get_model_name_from_path
from llava.eval.run_llava import eval_model

model_path = "mucai/vip-llava-7b"
prompt = "What is shown within the pointed region?"
image_file = "https://pages.cs.wisc.edu/~mucai/man-cross-street.jpg"

args = type('Args', (), {
    "model_path": model_path,
    "model_name": get_model_name_from_path(model_path),
    "query": prompt,
    "image_file": image_file,
    "conv_mode": None, "model_base": None, "temperature": 0.2, "top_p": None, "num_beams": 1, "max_new_tokens": 512, "sep": ",",
})()

eval_model(args)

Check out the details wth the load_pretrained_model function in llava/model/builder.py.

You can also use the eval_model function in llava/eval/run_llava.py to get the output easily. By doing so, you can use this code on Colab directly after downloading this repository.

ViP-LLaVA Weights

Please check out our Model Zoo for all public ViP-LLaVA checkpoints, and the instructions of how to use the weights.

Demo

To run our demo, you need to prepare LLaVA checkpoints locally. Please follow the instructions here to download the checkpoints.

Gradio Web UI

To launch a Gradio demo locally, please run the following commands one by one. If you plan to launch multiple model workers to compare between different checkpoints, you only need to launch the controller and the web server ONCE.

flowchart BT
    %% Declare Nodes
    gws("Gradio (UI Server)")
    c("Controller (API Server):<br/>PORT: 10000")
    mw7b("Model Worker:<br/>vip-llava-7b<br/>PORT: 40000")
    mw13b("Model Worker:<br/>vip-llava-13b<br/>PORT: 40001")

    %% Declare Styles
    classDef data fill:#3af,stroke:#48a,stroke-width:2px,color:#444
    classDef success fill:#8f8,stroke:#0a0,stroke-width:2px,color:#444
    classDef failure fill:#f88,stroke:#f00,stroke-width:2px,color:#444

    %% Assign Styles
    class id,od data;
    class cimg,cs_s,scsim_s success;
    class ncimg,cs_f,scsim_f failure;

    subgraph Demo Connections
        direction BT
        c<-->gws
        
        mw7b<-->c
        mw13b<-->c
    end

Launch a controller

python -m llava.serve.controller --host 0.0.0.0 --port 10000

Launch a gradio web server.

python -m llava.serve.gradio_web_server --controller http://localhost:10000 --model-list-mode reload

You just launched the Gradio web interface. Now, you can open the web interface with the URL printed on the screen. You may notice that there is no model in the model list. Do not worry, as we have not launched any model worker yet. It will be automatically updated when you launch a model worker.

Launch a model worker

This is the actual worker that performs the inference on the GPU. Each worker is responsible for a single model specified in --model-path.

python -m llava.serve.model_worker --host 0.0.0.0 --controller http://localhost:10000 --port 40000 --worker http://localhost:40000 --model-path mucai/vip-llava-13b

Wait until the process finishes loading the model and you see "Uvicorn running on ...". Now, refresh your Gradio web UI, and you will see the model you just launched in the model list.

You can launch as many workers as you want, and compare between different model checkpoints in the same Gradio interface. Please keep the --controller the same, and modify the --port and --worker to a different port number for each worker.

python -m llava.serve.model_worker --host 0.0.0.0 --controller http://localhost:10000 --port <different from 40000, say 40001> --worker http://localhost:<change accordingly, i.e. 40001> --model-path <ckpt2>

If you are using an Apple device with an M1 or M2 chip, you can specify the mps device by using the --device flag: --device mps.

Launch a model worker (Multiple GPUs, when GPU VRAM <= 24GB)

If the VRAM of your GPU is less than 24GB (e.g., RTX 3090, RTX 4090, etc.), you may try running it with multiple GPUs. Our latest code base will automatically try to use multiple GPUs if you have more than one GPU. You can specify which GPUs to use with CUDA_VISIBLE_DEVICES. Below is an example of running with the first two GPUs.

CUDA_VISIBLE_DEVICES=0,1 python -m llava.serve.model_worker --host 0.0.0.0 --controller http://localhost:10000 --port 40000 --worker http://localhost:40000 --model-path mucai/vip-llava-13b

Launch a model worker (4-bit, 8-bit inference, quantized)

You can launch the model worker with quantized bits (4-bit, 8-bit), which allows you to run the inference with reduced GPU memory footprint, potentially allowing you to run on a GPU with as few as 12GB VRAM. Note that inference with quantized bits may not be as accurate as the full-precision model. Simply append --load-4bit or --load-8bit to the model worker command that you are executing. Below is an example of running with 4-bit quantization.

python -m llava.serve.model_worker --host 0.0.0.0 --controller http://localhost:10000 --port 40000 --worker http://localhost:40000 --model-path mucai/vip-llava-13b --load-4bit

CLI Inference

Chat about images using ViP-LLaVA without the need of Gradio interface. It also supports multiple GPUs, 4-bit and 8-bit quantized inference. With 4-bit quantization, for our ViP-LLaVA-7B, it uses less than 8GB VRAM on a single GPU.

python -m llava.serve.cli \
    --model-path mucai/vip-llava-7b \
    --image-file "https://pages.cs.wisc.edu/~mucai/man-cross-street.jpg" \
    --load-4bit

Or use the bounding box as the

python -m llava.serve.cli_vip --model-path ./checkpoints/vip-llava-7b --image-file "https://pages.cs.wisc.edu/~mucai/example_styletransfer.png"  --bbox=100,200,200,300

Train

ViP-LLaVA training consists of three stages: (1) feature alignment stage: use our 558K subset of the LAION-CC-SBU dataset to connect a frozen pretrained vision encoder to a frozen LLM; (2) visual instruction tuning stage: 665K image-level instruction data from LLaVA-1.5 and 520K region-level instruction data using visual prompts. (3) finetuning on GPT-4V data.

LLaVA is trained on 8 A100 GPUs with 80GB memory. To train on fewer GPUs, you can reduce the per_device_train_batch_size and increase the gradient_accumulation_steps accordingly. Always keep the global batch size the same: per_device_train_batch_size x gradient_accumulation_steps x num_gpus.

Hyperparameters

We use a similar set of hyperparameters as Vicuna in finetuning. Both hyperparameters used in pretraining and finetuning are provided below.

Pretraining

Hyperparameter	Global Batch Size	Learning rate	Epochs	Max length	Weight decay
ViP-LLaVA-13B	256	1e-3	1	2048	0

Finetuning

Hyperparameter	Global Batch Size	Learning rate	Epochs	Max length	Weight decay
ViP-LLaVA-13B	128	2e-5	1	2048	0

Download Vicuna checkpoints (automatically)

Our base model Vicuna v1.5, which is an instruction-tuned chatbot, will be downloaded automatically when you run our provided training scripts. No action is needed.

Pretrain (feature alignment)

Please download the 558K subset of the LAION-CC-SBU dataset with BLIP captions we use in the paper here.

Pretrain takes around 5.5 hours for ViP-LLaVA-13B on 8x A100 (80G). It takes around 3.5 hours for ViP-LLaVA-7B.

Training script with DeepSpeed ZeRO-2: pretrain.sh.

--mm_projector_type mlp2x_gelu: the two-layer MLP vision-language connector.
--vision_tower clip_4layers_336: CLIP ViT-L/14 336px with multilayer feature fusion.

Pretrain takes around 20 hours for LLaVA-7B on 8x V100 (32G)

We provide training script with DeepSpeed here. Tips:

If you are using V100 which is not supported by FlashAttention, you can use the memory-efficient attention implemented in xFormers. Install xformers and replace llava/train/train_mem.py above with llava/train/train_xformers.py.

Visual Instruction Tuning

Prepare data

Please download the annotations of our instruction tuning data in stage 2 and stage 3. and download the images from constituting datasets:

COCO: train2017
GQA: images
OCR-VQA: download script, we save all files as .jpg
TextVQA: train_val_images
VisualGenome: part1, part2
Flickr30k: download here
VCR: download here
Visual7W: download here

After downloading all of them, organize the data as follows in ./playground/data,

├── flickr30k-images
├── v7w
├── vcr1images
├── coco
│   └── train2017
├── gqa
│   └── images
├── ocr_vqa
│   └── images
├── textvqa
│   └── train_images
└── vg
    ├── VG_100K
    └── VG_100K_2

Start training!

You may download our pretrained projectors in Model Zoo. It is not recommended to use legacy projectors, as they may be trained with a different version of the codebase, and if any option is off, the model will not function/train as we expected.

Visual instruction tuning takes around 40 hours for LLaVA-v1.5-13B on 8x A100 (80G) It takes around 20 hours for LLaVA-v1.5-7B on 8x A100 (40G).

Training script with DeepSpeed ZeRO-2: finetune_stage2.sh. If you further want to use GPT-4V data to enhance the chatting capbility, see the training script for stage 3 with DeepSpeed ZeRO-2: finetune_stage3.sh

If you are do not have enough GPU memory:

Use LoRA: finetune_lora.sh. We are able to fit 13B training in 8-A100-40G/8-A6000, and 7B training in 8-RTX3090. Make sure per_device_train_batch_size*gradient_accumulation_steps is the same as the provided script for best reproducibility.
Replace zero3.json with zero3_offload.json which offloads some parameters to CPU RAM. This slows down the training speed.

If you are interested in finetuning LLaVA model to your own task/data, please check out Finetune_Custom_Data.md。

Evaluation

ViP-LLaVA is both evaluate on the 4 academic region-level benchmark and the newly proposed ViP-Bench.

See Evaluation.md.

Citation

If you find ViP-LLaVA useful for your research and applications, please cite using this BibTeX:

@inproceedings{cai2024vipllava,
  author      = {Cai, Mu and Liu, Haotian and Mustikovela,  Siva Karthik and Meyer, Gregory P. and Chai, Yuning and Park, Dennis and Lee, Yong Jae},
  title       = {Making Large Multimodal Models Understand Arbitrary Visual Prompts},
  booktitle   = {IEEE Conference on Computer Vision and Pattern Recognition},
  year        = {2024}
}

Acknowledgement

Vicuna: the langauge model we built upon, and our base model Vicuna-13B that has the amazing language capabilities!
LLaVa: the codebase we built upon, which has amazing multimodal abalities!

vip-llava's People

Contributors

Stargazers

Watchers

Forkers

2132660698 tomchen-ctj tiantzhang anil2799 pzzhang saikoumonika sssssshf rayjue

vip-llava's Issues

[Question] Finetune Stage 2 Model

Question

First of all, great work and thank you so much for open-source it! I wonder if the stage 2 model(referred as ViP-LLaVA-Base) has been released anywhere? Maybe mucai/vip-llava-13b-pretrain? I am trying to finetune the stage 2 model using custom GPT instruction data. I am looking at scripts/finetune_stage3.sh and wonder if that's the correct script? But model used in the script is ./checkpoints/vip-llava-$model_size-stage2-ft and I dont really see it anywhere. Thank you!

Can u provide the subset only for ViP except llava mix original?

So that it can not be easily missleading model using llava original dataset.

Meanwhile, it looks like the images are missing..

    "id": "vcr-52941",
    "image": "vcr1images/lsmdc_3034_IDES_OF_MARCH/[email protected]",
    "meta_dir": "./dataset/vcr1images/lsmdc_3034_IDES_OF_MARCH/[email protected]",
    "class_names": [
        "person",

How to use visual prompts with huggingface?

Question

Thank you for the great work! What is the preferred way to feed in visual prompts with the huggingface endpoint?

[Question] Warning occurred when """model = LlavaLlamaForCausalLM.from_pretrained(model_path, low_cpu_mem_usage=True, **kwargs)""" in builder.py

Question

When I run bash scripts／finetune_stage3.sh or python predict.py, I faced the warning message. I have downloaded the pretrained weights from the huggingface.
Finetuning and Predicting can be done by ignoring this message. However, when I try to load finetuned weights(checkpoints／vip-llava-7b-stage3-ft), the same message occurs.
Is there any idea to sove this? Or can this message be ignored or not? Thanks in advance.

Warning Message

Some weights of the model checkpoint at ./model_weights were not used when initializing LlavaLlamaForCausalLM: ['model.vision_tower.vision_tower.vision_model.encoder.layers.12.self_attn.q_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.20.self_attn.v_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.12.layer_norm2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.8.self_attn.k_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.0.mlp.fc1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.7.self_attn.q_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.10.self_attn.v_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.12.mlp.fc2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.18.layer_norm1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.3.self_attn.q_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.17.mlp.fc2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.17.self_attn.out_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.5.self_attn.q_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.20.self_attn.k_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.1.self_attn.out_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.19.layer_norm1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.12.layer_norm2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.18.mlp.fc1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.8.self_attn.v_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.18.self_attn.q_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.8.layer_norm1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.19.layer_norm1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.7.mlp.fc1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.21.self_attn.q_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.13.mlp.fc1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.10.layer_norm1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.18.mlp.fc1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.7.self_attn.q_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.20.layer_norm2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.14.mlp.fc1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.23.mlp.fc2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.17.self_attn.out_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.14.layer_norm1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.2.self_attn.v_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.20.self_attn.q_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.2.self_attn.q_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.21.self_attn.q_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.16.self_attn.k_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.21.self_attn.v_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.13.layer_norm2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.10.self_attn.k_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.11.mlp.fc1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.16.layer_norm2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.10.layer_norm1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.1.mlp.fc1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.19.self_attn.v_proj.weight', 'model.vision_tower.vision_tower.vision_model.pre_layrnorm.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.19.self_attn.v_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.20.mlp.fc2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.18.layer_norm2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.0.self_attn.v_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.21.self_attn.k_proj.bias', 'model.vision_tower.vision_tower.vision_model.embeddings.class_embedding', 'model.vision_tower.vision_tower.vision_model.encoder.layers.1.self_attn.q_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.2.layer_norm1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.4.self_attn.v_proj.bias', 'model.vision_tower.vision_tower.vision_model.post_layernorm.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.12.self_attn.v_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.13.self_attn.v_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.0.self_attn.out_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.9.self_attn.out_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.23.layer_norm1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.16.mlp.fc2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.12.self_attn.k_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.20.mlp.fc2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.7.mlp.fc2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.0.self_attn.q_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.4.self_attn.k_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.18.layer_norm1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.19.mlp.fc1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.14.self_attn.v_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.2.mlp.fc1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.3.layer_norm1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.6.self_attn.out_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.14.mlp.fc1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.11.layer_norm1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.14.layer_norm1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.23.self_attn.out_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.22.self_attn.k_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.16.layer_norm1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.3.self_attn.q_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.6.self_attn.out_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.17.mlp.fc1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.18.self_attn.v_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.8.mlp.fc2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.5.mlp.fc1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.22.self_attn.q_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.5.self_attn.v_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.9.mlp.fc2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.14.self_attn.k_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.9.self_attn.v_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.11.self_attn.k_proj.weight', 'model.vision_tower.vision_tower.vision_model.embeddings.patch_embedding.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.1.self_attn.k_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.12.self_attn.out_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.8.self_attn.k_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.9.layer_norm1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.1.layer_norm2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.20.self_attn.k_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.17.self_attn.k_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.15.self_attn.v_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.16.layer_norm1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.23.self_attn.v_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.7.layer_norm1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.12.self_attn.q_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.4.layer_norm2.bias', 'model.vision_tower.vision_tower.vision_model.pre_layrnorm.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.21.self_attn.k_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.12.mlp.fc1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.2.layer_norm2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.1.self_attn.v_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.22.mlp.fc1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.3.layer_norm1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.10.mlp.fc1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.7.self_attn.k_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.7.layer_norm1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.2.self_attn.k_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.22.mlp.fc2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.18.mlp.fc2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.3.mlp.fc2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.13.self_attn.q_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.14.mlp.fc2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.11.mlp.fc2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.4.self_attn.v_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.9.self_attn.v_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.7.mlp.fc1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.11.self_attn.out_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.5.layer_norm2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.3.layer_norm2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.0.self_attn.k_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.16.mlp.fc1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.6.self_attn.v_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.8.self_attn.v_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.10.self_attn.out_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.2.mlp.fc2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.10.self_attn.k_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.12.layer_norm1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.21.layer_norm2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.17.self_attn.q_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.15.self_attn.out_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.19.self_attn.q_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.23.mlp.fc2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.2.layer_norm1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.18.self_attn.q_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.23.layer_norm2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.0.layer_norm2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.22.self_attn.k_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.8.self_attn.q_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.21.layer_norm1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.20.self_attn.v_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.3.self_attn.out_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.11.self_attn.q_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.14.mlp.fc2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.12.self_attn.v_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.23.self_attn.out_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.3.self_attn.k_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.4.self_attn.k_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.18.self_attn.k_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.9.mlp.fc2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.6.mlp.fc1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.6.self_attn.q_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.17.mlp.fc1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.14.self_attn.v_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.0.self_attn.out_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.22.layer_norm1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.19.self_attn.out_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.3.mlp.fc2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.18.self_attn.v_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.4.layer_norm1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.18.self_attn.out_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.4.self_attn.out_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.14.self_attn.q_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.7.self_attn.out_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.6.layer_norm2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.2.mlp.fc1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.11.self_attn.v_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.16.self_attn.v_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.14.self_attn.k_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.23.layer_norm1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.4.self_attn.q_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.1.self_attn.v_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.5.self_attn.out_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.16.self_attn.q_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.5.self_attn.k_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.6.self_attn.v_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.9.layer_norm2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.12.self_attn.out_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.8.mlp.fc1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.14.self_attn.out_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.15.mlp.fc1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.21.self_attn.out_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.19.self_attn.out_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.5.self_attn.out_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.17.self_attn.v_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.15.self_attn.q_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.10.mlp.fc2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.12.layer_norm1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.6.layer_norm1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.6.mlp.fc2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.10.mlp.fc1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.3.layer_norm2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.17.self_attn.q_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.4.layer_norm2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.19.self_attn.k_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.23.self_attn.k_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.0.self_attn.v_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.13.mlp.fc2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.19.mlp.fc2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.22.self_attn.v_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.15.self_attn.out_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.3.mlp.fc1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.0.layer_norm1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.11.self_attn.out_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.23.self_attn.q_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.4.mlp.fc2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.23.mlp.fc1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.21.self_attn.v_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.13.self_attn.k_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.19.self_attn.q_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.8.layer_norm1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.11.self_attn.v_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.7.self_attn.v_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.9.self_attn.out_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.4.layer_norm1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.8.mlp.fc1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.17.layer_norm1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.22.mlp.fc1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.23.self_attn.q_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.7.self_attn.v_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.2.layer_norm2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.2.self_attn.k_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.13.layer_norm2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.2.self_attn.out_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.19.mlp.fc2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.0.self_attn.k_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.14.self_attn.q_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.19.layer_norm2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.16.mlp.fc1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.12.mlp.fc2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.17.layer_norm2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.11.mlp.fc2.bias', 'model.vision_tower.vision_tower.vision_model.embeddings.position_embedding.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.15.layer_norm2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.3.self_attn.v_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.4.self_attn.q_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.22.self_attn.out_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.15.self_attn.k_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.18.layer_norm2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.18.mlp.fc2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.10.layer_norm2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.21.self_attn.out_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.5.self_attn.q_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.10.self_attn.out_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.12.mlp.fc1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.21.mlp.fc1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.13.self_attn.v_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.14.self_attn.out_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.15.self_attn.k_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.18.self_attn.out_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.23.mlp.fc1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.6.self_attn.k_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.19.mlp.fc1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.22.layer_norm2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.7.mlp.fc2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.11.layer_norm1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.5.self_attn.v_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.13.mlp.fc1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.9.self_attn.q_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.20.layer_norm2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.21.layer_norm1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.8.layer_norm2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.20.self_attn.q_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.1.self_attn.q_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.3.self_attn.k_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.10.layer_norm2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.8.self_attn.out_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.9.layer_norm2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.22.self_attn.out_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.0.layer_norm2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.10.self_attn.v_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.2.mlp.fc2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.22.self_attn.v_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.3.mlp.fc1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.0.self_attn.q_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.1.layer_norm1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.23.self_attn.k_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.16.self_attn.q_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.21.mlp.fc2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.10.self_attn.q_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.13.mlp.fc2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.22.layer_norm1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.22.layer_norm2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.11.layer_norm2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.10.mlp.fc2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.16.self_attn.k_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.6.layer_norm2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.3.self_attn.out_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.13.self_attn.out_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.2.self_attn.q_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.12.self_attn.k_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.15.layer_norm1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.15.layer_norm1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.21.mlp.fc2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.5.layer_norm1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.21.layer_norm2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.11.mlp.fc1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.18.self_attn.k_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.6.self_attn.k_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.19.layer_norm2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.23.layer_norm2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.17.layer_norm2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.1.mlp.fc1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.0.mlp.fc2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.1.mlp.fc2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.4.mlp.fc1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.7.layer_norm2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.1.mlp.fc2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.8.self_attn.out_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.20.self_attn.out_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.23.self_attn.v_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.6.mlp.fc2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.22.self_attn.q_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.13.self_attn.k_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.1.layer_norm1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.9.self_attn.k_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.5.layer_norm1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.20.mlp.fc1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.5.layer_norm2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.9.mlp.fc1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.15.mlp.fc2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.6.mlp.fc1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.3.self_attn.v_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.10.self_attn.q_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.16.self_attn.out_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.5.self_attn.k_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.8.layer_norm2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.17.mlp.fc2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.15.self_attn.q_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.14.layer_norm2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.21.mlp.fc1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.2.self_attn.v_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.17.self_attn.k_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.19.self_attn.k_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.4.mlp.fc1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.1.self_attn.k_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.5.mlp.fc2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.6.self_attn.q_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.2.self_attn.out_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.15.self_attn.v_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.7.self_attn.out_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.0.layer_norm1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.14.layer_norm2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.13.layer_norm1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.20.layer_norm1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.13.layer_norm1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.11.layer_norm2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.11.self_attn.k_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.9.self_attn.q_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.17.self_attn.v_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.15.mlp.fc1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.22.mlp.fc2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.9.self_attn.k_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.0.mlp.fc2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.4.mlp.fc2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.9.layer_norm1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.20.layer_norm1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.20.self_attn.out_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.5.mlp.fc1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.15.layer_norm2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.20.mlp.fc1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.8.self_attn.q_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.1.self_attn.out_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.17.layer_norm1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.9.mlp.fc1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.7.layer_norm2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.4.self_attn.out_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.13.self_attn.out_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.16.mlp.fc2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.16.layer_norm2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.5.mlp.fc2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.16.self_attn.out_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.16.self_attn.v_proj.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.11.self_attn.q_proj.weight', 'model.vision_tower.vision_tower.vision_model.post_layernorm.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.13.self_attn.q_proj.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.0.mlp.fc1.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.1.layer_norm2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.6.layer_norm1.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.15.mlp.fc2.bias', 'model.vision_tower.vision_tower.vision_model.encoder.layers.8.mlp.fc2.weight', 'model.vision_tower.vision_tower.vision_model.encoder.layers.7.self_attn.k_proj.weight']
- This IS expected if you are initializing LlavaLlamaForCausalLM from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).
- This IS NOT expected if you are initializing LlavaLlamaForCausalLM from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).

[Question] Can you provide register-specific instruction data with the help of GPT-4V?

Question

hi, author,
Thank you for your contribution. Could you please provide this dataset?

[Usage] pre-requisites for multi-image and multi-prompt generation

Describe the issue

Hi there, thank you for sharing this awesome project.

I have one question about the requried packages for multi-image and multi-prompt generation.
In the following link, it says that the version of the "transformers" library should be newer than 4.35.3.
(https://huggingface.co/llava-hf/vip-llava-7b-hf)

At first I installed all the necessary packages including transformers by following the instructions in a README file. (BTW, with the required versions of packages, I've got the same error discussed in #10 when I followed the steps in https://huggingface.co/llava-hf/vip-llava-7b-hf#using-pipeline )

In order to utilize the multi-image and prompt generation functions, I re-installed the transformers to satisfy the condition (4.36.0). However, when I tried to do an inference, there was a conflict between the transforemers and tokenizer libraries.

Could you share which version of packages you used when testing the multi-image and prompt generation options (transformers, tokenziers, accelerate, bitsandbytes and so on)?

ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
vip-llava 1.1.3 requires tokenizers<0.14,>=0.12.1, but you have tokenizers 0.15.2 which is incompatible.
vip-llava 1.1.3 requires transformers==4.31.0, but you have transformers 4.36.0 which is incompatible.

[Question] On the test set used for inference evaluation

Question

Hi, thanks for your contribution.
Could you please provide the test set link or converted file used in docs/Evaluation.md?

Training Stage

Question

Very exciting work, how long do “Visual Instruction Tuning“ and “Finetuning on GPT-4V Instruction Data“ take?

[Discussion] Unable to use the demo

Discussion

Thanks for your work, but the demo is hard to use, especially the module of 'Upload images'. Is there any solution to upload my own images successfully?

[Question] Will the code for creating Visual Instruction Tuning dataset be released?

Question

Waiting for Visual Instruction Tuning data, and i wonder if the code for generating this dataset will be released, thanks.

[Question] 请问这个文件怎么找不到了呢？convert_vipbench_for_eval.py

Question

请问这个文件怎么找不到了呢？convert_vipbench_for_eval.py

[Issue] Tensor mismatch at Runtime

RuntimeError: The expanded size of the tensor (628) must match the existing size (129) at non-singleton dimension 3. Target sizes: [1, 32, 1, 628]. Tensor sizes: [1, 1, 1, 129]

Issue: Upon running the demo colab notebook here, up until yesterday, I was able to get the expected results, but somehow since today the code throws the below mentioned error :

RuntimeError: The expanded size of the tensor (628) must match the existing size (129) at non-singleton dimension 3. Target sizes: [1, 32, 1, 628]. Tensor sizes: [1, 1, 1, 129]

The error gets triggered upon calling the following piece of code :

generate_ids = model.generate(**inputs, max_new_tokens=200)

Could you please have a look and possibly explain what causes the sudden mismatch?
Please note that the example instances used remain unchanged, i.e, the code is trhowing the tensor mismatch error even with the example included in the demo notebook.

[Question] How to get the VIP-Bench?

Question

Hello @mu-cai ,

Thanks for your excellent work. How can we get the VIP-Bench?

Release of Projector Weights

I want to fine-tune the ViP-LLaVA in my data. According to the instruction, I need to download the projector weight of the first stage, but it is displayed in MODEL_ZOO. Could you please release the projector weight? And I want to confirm the detailed steps for fine-tuning on custom datasets. Do I need to go through the Pretrain (feature alignment) stage from scratch? If not, can you publish the model weights after the Pretrain (feature alignment) stage? Thank you very much!

Looking forward to your reply!

[Question] When will the Visual Instruction Tuning data be released?

Question

Excellent work, may i ask when will the Visual Instruction Tuning data be released? Thx.

Recommend Projects

React

A declarative, efficient, and flexible JavaScript library for building user interfaces.
Vue.js

🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
Typescript

TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
TensorFlow

An Open Source Machine Learning Framework for Everyone
Django

The Web framework for perfectionists with deadlines.
Laravel

A PHP framework for web artisans
D3

Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

javascript

JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
web

Some thing interesting about web. New door for the world.
server

A server is a program made to process requests and deliver data to clients.
Machine learning

Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Visualization

Some thing interesting about visualization, use data art
Game

Some thing interesting about game, make everyone happy.

Recommend Org

Facebook

We are working to build community through open source technology. NB: members must have two-factor auth.
Microsoft

Open source projects and samples from Microsoft.
Google

Google ❤️ Open Source for everyone.
Alibaba

Alibaba Open Source for everyone
D3

Data-Driven Documents codes.
Tencent

China tencent open source team.

wisconsinaivision / vip-llava Goto Github PK

vip-llava's Introduction

🎯 [CVPR2024] ViP-LLaVA: Making Large Multimodal Models Understand Arbitrary Visual Prompts

Release

Contents

Install

Quick Start With HuggingFace

ViP-LLaVA Weights

Demo

Gradio Web UI

Launch a controller

Launch a gradio web server.

Launch a model worker

Launch a model worker (Multiple GPUs, when GPU VRAM <= 24GB)

Launch a model worker (4-bit, 8-bit inference, quantized)

CLI Inference

Train

Hyperparameters

Download Vicuna checkpoints (automatically)

Pretrain (feature alignment)

Visual Instruction Tuning

Evaluation

Citation

Acknowledgement

vip-llava's People

Contributors

Stargazers

Watchers

Forkers

vip-llava's Issues

Question

Question

Question

Question

Describe the issue

Question

Question

Discussion

Question

Question

RuntimeError: The expanded size of the tensor (628) must match the existing size (129) at non-singleton dimension 3. Target sizes: [1, 32, 1, 628]. Tensor sizes: [1, 1, 1, 129]

Question

Question

Recommend Projects

Recommend Topics

Recommend Org