metadata

license: apache-2.0
language:
  - en
pipeline_tag: image-text-to-text

Model description

xGen-MM is a series of the latest foundational Large Multimodal Models (LMMs) developed by Salesforce AI Research. This series advances upon the successful designs of the BLIP series, incorporating fundamental enhancements that ensure a more robust and superior foundation. These models have been trained at scale on high-quality image caption datasets and interleaved image-text data.

In the v1.5 (08/2024) release, we present a series of XGen-MM models including:

🤗 xGen-MM-instruct-interleave (our main instruct model): xgen-mm-phi3-mini-instruct-interleave-r-v1.5
- This model has higher overall scores than xGen-MM-instruct on both single-image and multi-image benchmarks.
🤗 xGen-MM-base: xgen-mm-phi3-mini-base-r-v1.5
🤗 xGen-MM-instruct: xgen-mm-phi3-mini-instruct-singleimg-r-v1.5
🤗 xGen-MM-instruct-dpo: xgen-mm-phi3-mini-instruct-dpo-r-v1.5

For more details, check out our tech report, fine-tuning code, and project page (coming soon).

DPO model results

Model	VLGuard (↓)	HallusionBench (↑)	POPE (↑)	MMBench (dev) (↑)	SEED-IMG (↑)	MMStar (↑)	MME (norm) (↑)
Phi-3-vision*	9.1	-	83.5	74.2	71.0	47.9	55.3
xgen-mm-phi3-mini-instruct-dpo-r-v1 (Ours)	5.2	56.6	86.8	76.4	72.1	47.1	64.4

(* = our eval)

We include some qualitative examples below of the safety features that complement our model's multimodal understanding capabilities.

How to use

Please check out our inference notebook for example code to use our model. We also provide an example script for batch inference.

Reproducibility:

Our evaluation is implemented based on open-compass/VLMEvalKit. We will create a PR to that repo to support XGen-MM evaluation.

Bias, Risks, Limitations, and Ethical Considerations

The main data sources are from the internet, including webpages, image stock sites, and curated datasets released by the research community. We have excluded certain data, such as LAION, due to known CSAM concerns. The model may be subject to bias from the original data source, as well as bias from LLMs and commercial APIs. We strongly recommend users assess safety and fairness before applying to downstream applications.

License

Our code and weights are released under the Apache 2.0 license.

Code acknowledgment

Our training code is based on OpenFlamingo: An open-source framework for training large multimodal models., and part of our data preprocessing code is adapted from LLaVA. Our evaluation code is based on VLMEvalKit: Open-source evaluation toolkit of large vision-language models (LVLMs).

We thank the authors for their open-source implementations.

Citation

@misc{blip3-xgenmm,
  author          = {Le Xue, Manli Shu, Anas Awadalla, Jun Wang, An Yan, Senthil Purushwalkam, Honglu Zhou, Viraj Prabhu, Yutong Dai, Michael S Ryoo, Shrikant Kendre, Jieyu Zhang, Can Qin, Shu Zhang, Chia-Chih Chen, Ning Yu, Juntao Tan, Tulika Manoj Awalgaonkar, Shelby Heinecke, Huan Wang, Yejin Choi, Ludwig Schmidt, Zeyuan Chen, Silvio Savarese, Juan Carlos Niebles, Caiming Xiong, Ran Xu},
  title           = {xGen-MM (BLIP-3): A Family of Open Large Multimodal Models},
  year            = {2024},
  eprint          = {2408.08872},
  archivePrefix   = {arXiv},
  primaryClass    = {cs.CV},
  url             = {https://arxiv.org/abs/2408.08872}, 
}

Troubleshoot

If you missed any packages, please consider the following

pip install torch==2.2.1 torchvision==0.17.1 torchaudio==2.2.1 --index-url https://download.pytorch.org/whl/cu121
pip install open_clip_torch==2.24.0
pip install einops
pip install einops-exts
pip install transformers==4.41.1