Name	Name	Last commit message	Last commit date
parent directory ..
eval	eval
internvl	internvl
shell	shell
tools	tools
README.md	README.md
evaluate.sh	evaluate.sh
pyproject.toml	pyproject.toml
zero_stage1_config.json	zero_stage1_config.json
zero_stage2_config.json	zero_stage2_config.json
zero_stage3_config.json	zero_stage3_config.json

InternVL-Chat

This folder contains the implementation of the InternVL-Chat.

🛠️ Installation

In addition, using this codebase requires executing the following steps:

Install other requirements:

pip install --upgrade pip  # enable PEP 660 support
pip install -e .

📦 Model Preparation

model name	type	download	size
InternViT-300M-448px	ViT	🤗 HF link	0.6 GB
InternViT-6B-448px-V1-2	ViT	🤗 HF link	11.1 GB
InternViT-6B-448px-V1-5	ViT	🤗 HF link	11.1 GB
Nous-Hermes-2-Yi-34B	LLM	🤗 HF link	65.0 GB

Please download the above model weights and place them in the pretrained/ folder.

cd pretrained/
# pip install -U huggingface_hub
huggingface-cli download --resume-download --local-dir-use-symlinks False OpenGVLab/InternViT-300M-448px --local-dir InternViT-300M-448px
huggingface-cli download --resume-download --local-dir-use-symlinks False OpenGVLab/InternViT-6B-448px-V1-2 --local-dir InternViT-6B-448px-V1-2
huggingface-cli download --resume-download --local-dir-use-symlinks False OpenGVLab/InternViT-6B-448px-V1-5 --local-dir InternViT-6B-448px-V1-5
huggingface-cli download --resume-download --local-dir-use-symlinks False NousResearch/Nous-Hermes-2-Yi-34B --local-dir Nous-Hermes-2-Yi-34B

The directory structure is:

pretrained
│── InternViT-300M-448px/
│── InternViT-6B-448px-V1-2/
│── InternViT-6B-448px-V1-5/
└── Nous-Hermes-2-Yi-34B/

🔥 Supervised Fine-tuning

Prepare Training Datasets

Inspired by LLaVA-NeXT, we adopted a data-efficient SFT strategy to train InternVL-Chat-V1-2, utilizing approximately 1.2M of visual instruction tuning samples in total, all of which are fully open-source. In a macro sense, we build upon ShareGPT-4V and additionally integrate LLaVA-ZH, DVQA, ChartQA, AI2D, DocVQA, GeoQA+, and SynthDoG-EN. Most of the data remains consistent with LLaVA-NeXT.

First, download the annotation files and place them in the playground/opensource/ folder.

Second, download all the images we used.

AI2D: ai2d_images (provided by InternLM-XComposer)
ChartQA: ChartQA Dataset
COCO: train2017
DocVQA: train, val, test
DVQA: images
GQA: images
LLaVA-Pretrain: images
OCR-VQA: download script. We save all files as .jpg
SAM: We only use 000000~000050.tar for now. You can quickly download 9K images from here.
TextVQA: trainvalimages
SynthDoG-EN: We only use 00000~00004 parquet files for now, with a total of 30K images. We provide the converted images.
VisualGenome: part1, part2
WebData: images. Only for academic usage.
GeoQA+: GeoQA+ images

Then, organize the data as follows in playground/data:

playground/
├── opensource
│   ├── ai2d_train_12k.jsonl
│   ├── chartqa_train_18k.jsonl
│   ├── docvqa_train_10k.jsonl
│   ├── dvqa_train_200k.jsonl
│   ├── geoqa+.jsonl
│   ├── llava_instruct_150k_zh.jsonl
│   ├── sharegpt4v_instruct_gpt4-vision_cap100k.jsonl
│   ├── sharegpt4v_mix665k_cap23k_coco-ap9k_lcs3k_sam9k_div2k.jsonl
│   └── synthdog_en.jsonl
├── data
│   ├── ai2d
│   │   ├── abc_images
│   │   └── images
│   ├── chartqa
│   │   ├── test
│   │   ├── train
│   │   └── val
│   ├── coco
│   │   └── train2017
│   ├── docvqa
│   │   ├── test
│   │   ├── train
│   │   └── val
│   ├── dvqa
│   │   └── images
│   ├── gqa
│   │   └── images
│   ├── llava
│   │   └── llava_pretrain
│   │       └── images
│   ├── ocr_vqa
│   │   └── images
│   ├── sam
│   │   └── images
│   ├── share_textvqa
│   │   └── images
│   ├── synthdog-en
│   │   └── images
│   ├── textvqa
│   │   └── train_images
│   ├── vg
│   │   ├── VG_100K
│   │   └── VG_100K_2
│   ├── web-celebrity
│   │   └── images
│   ├── web-landmark
│   │   └── images
│   ├── wikiart
│   │   └── images
│   ├── geoqa+
│   │   └── images

Start Training

We provide slurm scripts for multi-node multi-GPU training. You can use either 32 or 64 GPUs to train this model. If you use 64 GPUs, training will take approximately 18 hours.

If you encounter an OOM error, you can decrease the PER_DEVICE_BATCH_SIZE, for example, set PER_DEVICE_BATCH_SIZE=4.

# using 32 GPUs
PARTITION='your partition' GPUS=32 PER_DEVICE_BATCH_SIZE=8 sh shell/hermes2_yi34b/internvl_chat_v1_2_hermes2_yi34b_448_res_finetune.sh
# using 64 GPUs
PARTITION='your partition' GPUS=64 PER_DEVICE_BATCH_SIZE=8 sh shell/hermes2_yi34b/internvl_chat_v1_2_hermes2_yi34b_448_res_finetune.sh

The hyperparameters used for fine-tuning are listed in the following table. And, you can view the training logs in tensorboard at here.

Hyperparameter	Trainable Param	Global Batch Size	Learning rate	Epochs	Max length	Weight decay
InternVL-Chat-V1-2	40B (full model)	512	1e-5	1	2048	0.05

Continued Fine-tune

See this document to finetune InternVL-Chat-V1-2.

📊 Evaluation

OCR-related Benchmarks

Note: TextVQA contains two scores, representing not using or using Rosetta OCR tokens, respectively.

model	#param	DocVQA (val/test)	ChartVQA (avg. test)	InfoVQA (val/test)	TextVQA (val, wo/w OCR)	OCRBench	AI2D
InternVL‑Chat‑V1‑1	19B	47.6 / 48.1	59.9	33.3 / 32.0	64.2 / 68.6	530	72.4
InternVL‑Chat‑V1‑2	40B	56.4 / 57.7	68.0	36.0 / 39.5	67.5 / 72.5	569	79.0
InternVL‑Chat‑V1‑2‑Plus	40B	56.9 / 56.8	72.8	40.9 / 40.6	71.2 / 74.1	598	78.9
InternVL‑Chat‑V1‑5	26B	90.5 / 90.8	83.8	72.4 / 72.5	80.6 / -	724	80.7

MultiModal Benchmark

model	#param	MME	MMB (dev/test)	MMB‑CN (dev/test)	CCBench	MMVet	MMMU (val/test)	MathVista (testmini)	Hallusion Bench	RealWorld QA	SEEDv1 (image)	CMMMU (val/test)	POPE	MMVP	Tiny LVLM	LLaVA Wild
InternVL‑Chat‑V1‑1	19B	1659.8 / 361.4	76.7 / 75.4	71.9 / 70.3	43.3	46.7	39.1 / 35.3	34.5	36.1	58.0	73.2	34.8 / 34.0	87.1	44.7	343.2	73.2
InternVL‑Chat‑V1‑2	40B	1686.8 / 488.6	81.4 / 82.2	79.5 / 81.2	58.6	48.9	51.6 / 46.2	47.7	47.6	67.5	75.6	-	88.0	56.7	350.3	85.0
InternVL‑Chat‑V1‑2‑Plus	40B	1625.2 / 552.9	83.4 / 83.8	81.6 / 82.0	55.9	47.9	50.3 / 45.6	59.9	47.4	67.8	76.4	-	88.7	58.7	353.9	84.6
InternVL‑Chat‑V1‑5	26B	1637.8 / 550.0	- / 82.2	- / 82.0	70.0	62.8	45.2 / -	53.5	49.3	66.0	76.0	-	88.3	57.3	356.8	94.7

Visual Question Answering & Image Captioning

model	#param	OKVQA (val)	VizWiz (val/test)	GQA (test)	SQA (image)	VQAv2 (testdev)	COCO (test)	Flickr30K (test)	NoCaps (val)
InternVL‑Chat‑V1‑1	19B	64.1	59.0 / 57.3	62.5	90.1	80.9	142.2	84.8	120.8
InternVL‑Chat‑V1‑2	40B	62.5	61.9 / 60.0	64.0	83.3	-	113.9	92.9	112.5
InternVL‑Chat‑V1‑2‑Plus	40B	67.6	61.3 / 59.5	66.9	98.1	-	143.4	89.5	125.8
InternVL‑Chat‑V1‑5	26B	62.0	63.5 / -	65.7	94.0	-	98.4	81.2	99.6

Visual Grounding

model	#param	RefCOCO (val)	RefCOCO (testA)	RefCOCO (testB)	RefCOCO+ (val)	RefCOCO+ (testA)	RefCOCO+ (testB)	RefCOCO‑g (val)	RefCOCO‑g (test)
InternVL‑Chat‑V1‑1	19B	84.7	89.9	78.6	78.5	85.6	70.1	81.0	81.4
InternVL‑Chat‑V1‑2	40B	74.4	80.3	66.5	70.7	77.6	62.0	69.2	70.0
InternVL‑Chat‑V1‑2‑Plus	40B	90.2	93.4	85.5	85.3	90.4	79.7	88.5	88.8
InternVL‑Chat‑V1‑5	26B	91.4	93.7	87.1	87.0	92.3	80.9	88.5	89.3

📊 Evaluation (Legacy Models)

model	QLLaMA	LLM	res	COCO	Flickr	NoCaps	VQAv2	GQA	VizWiz	TextVQA	MME	POPE	Download
InternVL‑Chat	✔️	frozen V‑7B	224	141.4	89.7	120.5	72.3	57.7	44.5	42.1	1298.5	85.2	TODO
InternVL‑Chat	✔️	frozen V‑13B	224	142.4	89.9	123.1	71.7	59.5	54.0	49.1	1317.2	85.4	TODO
InternVL‑Chat	✔️	V‑13B	336	146.2	92.2	126.2	81.2	66.6	58.5	61.5	1586.4	87.6	TODO

❓ How to Evaluate

Image Caption Benchmarks

COCO Karpathy test

COCO images are used in VQAv2/OK-VQA/RefCOCO/RefCOCO+/RefCOCOg. Make sure you have already downloaded COCO images before evaluating on these benchmarks.

Data Preparation

mkdir -p data/coco && cd data/coco

# download coco images
wget http://images.cocodataset.org/zips/train2014.zip && unzip train2014.zip
wget http://images.cocodataset.org/zips/val2014.zip && unzip val2014.zip
wget http://images.cocodataset.org/zips/test2015.zip && unzip test2015.zip

mkdir -p annotations && cd annotations/
# download converted annotation files
wget https://github.com/OpenGVLab/InternVL/releases/download/data/coco_karpathy_test.json
wget https://github.com/OpenGVLab/InternVL/releases/download/data/coco_karpathy_test_gt.json

cd ../../../

Evaluation

GPUS=8 sh evaluate.sh <checkpoint> caption-coco [--dynamic]

Flickr30K Karpathy test

Data Preparation

mkdir -p data/flickr30k && cd data/flickr30k

# download images from https://bryanplummer.com/Flickr30kEntities/
# karpathy split annotations can be downloaded from the following link:
# https://github.com/mehdidc/retrieval_annotations/releases/download/1.0.0/flickr30k_test_karpathy.txt
# this file is provided by the clip-benchmark repository.
# We convert this txt file to json format, download the converted file:
wget https://github.com/OpenGVLab/InternVL/releases/download/data/flickr30k_test_karpathy.json

cd ../..

Evaluation

GPUS=8 sh evaluate.sh <checkpoint> caption-flickr30k [--dynamic]

NoCaps val

Data Preparation

mkdir -p data/nocaps && cd data/nocaps

# download images from https://nocaps.org/download
# original annotations can be downloaded from https://nocaps.s3.amazonaws.com/nocaps_val_4500_captions.json
wget https://nocaps.s3.amazonaws.com/nocaps_val_4500_captions.json

cd ../..

Evaluation

GPUS=8 sh evaluate.sh <checkpoint> caption-nocaps [--dynamic]

General VQA Benchmarks

VQAv2 val & test-dev

Data Preparation

mkdir -p data/vqav2 && cd data/vqav2

# make sure you have downloaded COCO images
ln -s ../coco/train2014 ./
ln -s ../coco/val2014 ./
ln -s ../coco/test2015 ./

# download questions and annotations
wget https://s3.amazonaws.com/cvmlp/vqa/mscoco/vqa/v2_Annotations_Train_mscoco.zip && unzip v2_Annotations_Train_mscoco.zip
wget https://s3.amazonaws.com/cvmlp/vqa/mscoco/vqa/v2_Questions_Train_mscoco.zip && unzip v2_Questions_Train_mscoco.zip
wget https://s3.amazonaws.com/cvmlp/vqa/mscoco/vqa/v2_Annotations_Val_mscoco.zip && unzip v2_Annotations_Val_mscoco.zip
wget https://s3.amazonaws.com/cvmlp/vqa/mscoco/vqa/v2_Questions_Val_mscoco.zip && unzip v2_Questions_Val_mscoco.zip
wget https://s3.amazonaws.com/cvmlp/vqa/mscoco/vqa/v2_Questions_Test_mscoco.zip && unzip v2_Questions_Test_mscoco.zip

# download converted files
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/vqav2/vqav2_train.jsonl
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/vqav2/vqav2_val.jsonl
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/vqav2/vqav2_testdev.jsonl

cd ../..

Evaluation

# VQAv2-val
GPUS=8 sh evaluate.sh <checkpoint> vqa-vqav2-val [--dynamic]
# VQAv2-testdev
GPUS=8 sh evaluate.sh <checkpoint> vqa-vqav2-testdev [--dynamic]

For the testdev set, submit the results to the evaluation server.

OKVQA val

Data Preparation

mkdir -p data/okvqa && cd data/okvqa

# make sure you have downloaded COCO images
ln -s ../coco/train2014 ./
ln -s ../coco/val2014 ./

# download annotations and questions
wget https://okvqa.allenai.org/static/data/mscoco_train2014_annotations.json.zip && unzip mscoco_train2014_annotations.json.zip
wget https://okvqa.allenai.org/static/data/OpenEnded_mscoco_train2014_questions.json.zip && unzip OpenEnded_mscoco_train2014_questions.json.zip
wget https://okvqa.allenai.org/static/data/mscoco_val2014_annotations.json.zip && unzip mscoco_val2014_annotations.json.zip
wget https://okvqa.allenai.org/static/data/OpenEnded_mscoco_val2014_questions.json.zip && unzip OpenEnded_mscoco_val2014_questions.json.zip

# download converted files
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/okvqa/okvqa_train.jsonl
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/okvqa/okvqa_val.jsonl

cd ../..

Evaluation

GPUS=8 sh evaluate.sh <checkpoint> vqa-okvqa-val [--dynamic]

TextVQA val

Data Preparation

mkdir -p data/textvqa && cd data/textvqa

# download images
wget https://dl.fbaipublicfiles.com/textvqa/images/train_val_images.zip && unzip train_val_images.zip

# download converted files
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/textvqa/textvqa_train_annotations.json
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/textvqa/textvqa_train_questions.json
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/textvqa/textvqa_train.jsonl
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/textvqa/textvqa_val_annotations.json
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/textvqa/textvqa_val_questions.json
wget https://huggingface.co/OpenGVLab/InternVL/raw/main/textvqa_val.jsonl
wget https://huggingface.co/OpenGVLab/InternVL/raw/main/textvqa_val_llava.jsonl

cd ../..

Evaluation

# without ocr tokens
GPUS=8 sh evaluate.sh <checkpoint> vqa-textvqa-val [--dynamic]
# with ocr tokens (hint: LLaVA use ocr tokens)
GPUS=8 sh evaluate.sh <checkpoint> vqa-textvqa-val-ocr [--dynamic]

VizWiz val & test

Data Preparation

mkdir -p data/vizwiz && cd data/vizwiz

# download images
wget https://vizwiz.cs.colorado.edu/VizWiz_final/images/train.zip && unzip train.zip
wget https://vizwiz.cs.colorado.edu/VizWiz_final/images/val.zip && unzip val.zip
wget https://vizwiz.cs.colorado.edu/VizWiz_final/images/test.zip && unzip test.zip

# download annotations
wget https://vizwiz.cs.colorado.edu/VizWiz_final/vqa_data/Annotations.zip && unzip Annotations.zip

# download converted files
# train
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/vizwiz/vizwiz_train_annotations.json
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/vizwiz/vizwiz_train_questions.json
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/vizwiz/vizwiz_train.jsonl
# val
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/vizwiz/vizwiz_val_annotations.json
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/vizwiz/vizwiz_val_questions.json
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/vizwiz/vizwiz_val.jsonl
# test
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/vizwiz/vizwiz_test.jsonl
cd ../..

Evaluation

# VizWiz val
GPUS=8 sh evaluate.sh <checkpoint> vqa-vizwiz-val [--dynamic]
# VizWiz test
GPUS=8 sh evaluate.sh <checkpoint> vqa-vizwiz-test [--dynamic]

For the test set, submit the results to the evaluation server.

DocVQA val & test

Data Preparation

mkdir -p data/docvqa && cd data/docvqa

# download images and annotations
wget https://datasets.cvc.uab.es/rrc/DocVQA/train.tar.gz --no-check-certificate # (optional)
wget https://datasets.cvc.uab.es/rrc/DocVQA/val.tar.gz --no-check-certificate
wget https://datasets.cvc.uab.es/rrc/DocVQA/test.tar.gz --no-check-certificate

# unzip files
tar -zxvf train.tar.gz
tar -zxvf val.tar.gz
tar -zxvf test.tar.gz

# download converted jsonl files
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/docvqa/train.jsonl
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/docvqa/val.jsonl
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/docvqa/test.jsonl
cd ../..

Evaluation

# DocVQA-val
GPUS=8 sh evaluate.sh <checkpoint> vqa-docvqa-val [--dynamic]
# DocVQA-test
GPUS=8 sh evaluate.sh <checkpoint> vqa-docvqa-test [--dynamic]

For the test set, submit the results to the evaluation server.

ChartQA test-human & test-augmented

Data Preparation

mkdir -p data/chartqa && cd data/chartqa

# download images from https://drive.google.com/file/d/1Lm_w6zeET1Hyl_9ks6w5nEsgpoyPHalV/view

# download converted files
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/chartqa/train_human.jsonl
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/chartqa/train_augmented.jsonl
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/chartqa/test_human.jsonl
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/chartqa/test_augmented.jsonl

cd ../..

Evaluation

# test both ChartQA-test-human & ChartQA-test-augmented
GPUS=8 sh evaluate.sh <checkpoint> vqa-chartqa-test [--dynamic]

GQA testdev

Data Preparation

mkdir -p data/gqa && cd data/gqa

# download images
wget https://downloads.cs.stanford.edu/nlp/data/gqa/images.zip
unzip images.zip

# download converted files
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/gqa/testdev_balanced.jsonl
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/gqa/train_balanced.jsonl
wget https://github.com/OpenGVLab/InternVL/releases/download/data/llava_gqa_testdev_balanced_qwen_format.jsonl

cd ../..

Evaluation

GPUS=8 sh evaluate.sh <checkpoint> vqa-gqa-testdev [--dynamic]

OCRVQA val & test

Data Preparation

mkdir -p data/ocrvqa && cd data/ocrvqa

# download images by following instructions at https://ocr-vqa.github.io/kvqa_ProjectFiles/README.txt

# download converted files
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/ocrvqa/ocrvqa_train.jsonl
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/ocrvqa/ocrvqa_val.jsonl
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/ocrvqa/ocrvqa_test.jsonl

cd ../..

Evaluation

# OCRVQA-val
GPUS=8 sh evaluate.sh <checkpoint> vqa-ocrvqa-val [--dynamic]
# OCRVQA-test
GPUS=8 sh evaluate.sh <checkpoint> vqa-ocrvqa-test [--dynamic]

AI2D test

Data Preparation

mkdir -p data/ai2diagram && cd data/ai2diagram
# download converted files
wget https://huggingface.co/OpenGVLab/InternVL/raw/main/ai2d_test_vlmevalkit.jsonl -O test_vlmevalkit.jsonl
wget https://huggingface.co/OpenGVLab/InternVL/resolve/main/AI2D_TEST.zip && unzip AI2D_TEST.zip

# download images from Google drive (optional, provided by InternLM-XComposer)
# https://drive.google.com/file/d/1dqqa3MnrxMXaU_K9JA6C83je32ibwdOY/view?usp=sharing
# images should be placed in `data/ai2diagram/ai2d/abc_images` and `data/ai2diagram/ai2d/images`
cd ../..

Evaluation

GPUS=8 sh evaluate.sh <checkpoint> vqa-ai2d-test [--dynamic]

ScienceQA test

Data Preparation

mkdir -p data/scienceqa/images && cd data/scienceqa/images

# download images
wget https://scienceqa.s3.us-west-1.amazonaws.com/images/test.zip && unzip test.zip

cd ..

# download original questions
wget https://github.com/lupantech/ScienceQA/blob/main/data/scienceqa/problems.json

# download converted files
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/scienceqa/scienceqa_test_img.jsonl

cd ../..

Evaluation

GPUS=8 sh evaluate.sh <checkpoint> scienceqa [--dynamic]

Refer Expression Comprehension

RefCOCO/RefCOCO+/RefCOCO-g

Data Preparation

mkdir -p data/refcoco && cd data/refcoco

# download converted files
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/refcoco/refcoco_val.jsonl
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/refcoco/refcoco_testA.jsonl
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/refcoco/refcoco_testB.jsonl
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/refcoco%2B/refcoco%2B_val.jsonl
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/refcoco%2B/refcoco%2B_testA.jsonl
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/refcoco%2B/refcoco%2B_testB.jsonl
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/refcocog/refcocog_val.jsonl
wget https://ofasys-wlcb.oss-cn-wulanchabu.aliyuncs.com/Qwen-VL/evaluation/refcocog/refcocog_test.jsonl

cd ../..

Evaluation

GPUS=8 sh evaluate.sh <checkpoint> refcoco [--dynamic]

MultiModal Benchmarks

MME

Data Preparation

mkdir -p data/mme && cd data/mme

# 1. Download the data following the official instructions [here](https://github.com/BradyFU/Awesome-Multimodal-Large-Language-Models/tree/Evaluation).
# 2. Downloaded images to `MME_Benchmark_release_version`.

cd ../..

Evaluation

# single GPU testing
CUDA_VISIBLE_DEVICES=0 sh evaluate.sh <checkpoint> mme

MMBench dev & test

Data Preparation

mkdir -p data/mmbench && cd data/mmbench

# download csv files of mmbench
wget http://opencompass.openxlab.space/utils/MMBench/CCBench_legacy.tsv
wget https://download.openmmlab.com/mmclassification/datasets/mmbench/mmbench_dev_20230712.tsv
wget https://download.openmmlab.com/mmclassification/datasets/mmbench/mmbench_dev_cn_20231003.tsv
wget https://download.openmmlab.com/mmclassification/datasets/mmbench/mmbench_dev_en_20231003.tsv
wget https://download.openmmlab.com/mmclassification/datasets/mmbench/mmbench_test_cn_20231003.tsv
wget https://download.openmmlab.com/mmclassification/datasets/mmbench/mmbench_test_en_20231003.tsv

cd ../..

Evaluation

# mmbench_dev_20230712
GPUS=8 sh evaluate.sh <checkpoint> mmbench-dev-en [--dynamic]
# mmbench_dev_cn_20231003
GPUS=8 sh evaluate.sh <checkpoint> mmbench-dev-cn [--dynamic]
# mmbench_test_en_20231003
GPUS=8 sh evaluate.sh <checkpoint> mmbench-test-en [--dynamic]
# mmbench_test_cn_20231003
GPUS=8 sh evaluate.sh <checkpoint> mmbench-test-cn [--dynamic]
# ccbench_dev
GPUS=8 sh evaluate.sh <checkpoint> ccbench-dev [--dynamic]

Then, submit the results to the evaluation server.

POPE

Data Preparation

mkdir -p data/pope && cd data/pope

# make sure you have downloaded COCO images
ln -s ../coco/val2014 ./
wget https://github.com/OpenGVLab/InternVL/releases/download/data/llava_pope_test.jsonl

# download `coco` from POPE
mkdir -p coco && cd coco
wget https://github.com/AoiDragon/POPE/raw/e3e39262c85a6a83f26cf5094022a782cb0df58d/output/coco/coco_pope_adversarial.json
wget https://github.com/AoiDragon/POPE/raw/e3e39262c85a6a83f26cf5094022a782cb0df58d/output/coco/coco_pope_popular.json
wget https://github.com/AoiDragon/POPE/raw/e3e39262c85a6a83f26cf5094022a782cb0df58d/output/coco/coco_pope_random.json
cd ../../..

Evaluation

GPUS=8 sh evaluate.sh <checkpoint> pope [--dynamic]

MMMU

Data Preparation

The evaluation code will automatically download the dataset from hugging face.

Evaluation

# dev set
GPUS=8 sh evaluate.sh <checkpoint> mmmu-dev [--dynamic]
# val set
GPUS=8 sh evaluate.sh <checkpoint> mmmu-val [--dynamic]
# test set
GPUS=8 sh evaluate.sh <checkpoint> mmmu-test [--dynamic]

For the test set, submit the results to the evaluation server.

Tiny LVLM

Data Preparation

mkdir -p data/tiny_lvlm && cd data/tiny_lvlm

# download dataset from https://github.com/OpenGVLab/Multi-Modality-Arena/tree/main/tiny_lvlm_evaluation
# i.e., download `updated_datasets.tar.gz` from https://drive.google.com/file/d/1PuFC612XzOmKwzRldtBb1CFZnIjiR7we/view
tar -xzvf updated_datasets.tar.gz

cd ../..

Evaluation

GPUS=8 sh evaluate.sh <checkpoint> tiny_lvlm [--dynamic]

LLaVA Bench

Data Preparation

cd data/
# download dataset from https://huggingface.co/datasets/liuhaotian/llava-bench-in-the-wild
git clone https://huggingface.co/datasets/liuhaotian/llava-bench-in-the-wild
cd llava-bench-in-the-wild/
rm -rf images && mkdir -p images && cd images
# download all 24 images
wget https://huggingface.co/datasets/liuhaotian/llava-bench-in-the-wild/resolve/main/images/001.jpg
# ...
wget https://huggingface.co/datasets/liuhaotian/llava-bench-in-the-wild/resolve/main/images/024.jpg
cd ../../../

Evaluation

# single GPU testing
export OPENAI_API_KEY='your_gpt4_key'
CUDA_VISIBLE_DEVICES=0 sh evaluate.sh <checkpoint> llava-bench

MM-Vet

Data Preparation

mkdir -p data/mm-vet && cd data/mm-vet
wget https://github.com/yuweihao/MM-Vet/releases/download/v1/mm-vet.zip
unzip mm-vet.zip
wget https://huggingface.co/OpenGVLab/InternVL/raw/main/llava-mm-vet.jsonl
cd ../..

Evaluation

GPUS=8 sh evaluate.sh <checkpoint> mmvet [--dynamic]

MMVP

Data Preparation

cd data
git lfs install
git clone https://huggingface.co/datasets/MMVP/MMVP
git clone https://huggingface.co/datasets/MMVP/MMVP_VLM
cd ..

Evaluation

GPUS=8 sh evaluate.sh <checkpoint> mmvp [--dynamic]

MathVista

Data Preparation

mkdir -p data/MathVista && cd data/MathVista
wget https://huggingface.co/datasets/AI4Math/MathVista/raw/main/annot_testmini.json
cd ../..

Evaluation

export OPENAI_API_KEY='your-openai-key'
# testmini set
GPUS=8 sh evaluate.sh <checkpoint> mathvista-testmini [--dynamic]
# test set
GPUS=8 sh evaluate.sh <checkpoint> mathvista-test [--dynamic]

SEED

Data Preparation

mkdir -p data/SEED && cd data/SEED
# 1. Follow the official instructions [Data Preparation for SEED-Bench-1](https://github.com/AILab-CVC/SEED-Bench/blob/main/DATASET.md#data-preparation-for-seed-bench-1)
#    to download the images and the videos. Put images under `./data/SEED/SEED-Bench-image`.
# 2. Extract the video frame in the middle from the downloaded videos, and put them under `./data/SEED/SEED-Bench-image`.
#    LLaVA provided the script [`extract_video_frames.py`](../internvl_chat/tools/extract_video_frames.py) modified from the official one.

wget https://huggingface.co/OpenGVLab/InternVL/raw/main/seed.jsonl
cd ../..

Evaluation

GPUS=8 sh evaluate.sh <checkpoint> seed [--dynamic]

Files

internvl_chat

Directory actions

More options

Directory actions

More options

Latest commit

History

internvl_chat

Folders and files

parent directory

InternVL-Chat

🛠️ Installation

📦 Model Preparation

🔥 Supervised Fine-tuning

Prepare Training Datasets

Start Training

Continued Fine-tune

📊 Evaluation

📊 Evaluation (Legacy Models)

❓ How to Evaluate

Image Caption Benchmarks

General VQA Benchmarks

Refer Expression Comprehension

RefCOCO/RefCOCO+/RefCOCO-g

MultiModal Benchmarks