--- license: cc-by-nc-4.0 language: - en pipeline_tag: image-text-to-text --- # Model description `BLIP3` is a series of foundational vision-language models (VLMs) developed by Salesforce AI Research. \ These models have been trained at scale on high-quality image caption datasets and interleaved image-text data. BLIP3 highlights a few features below, * The pretrained foundation model, `blip3-phi3-mini-base-r-v1`, achieves state-of-the-art performance under 5b parameters and demonstrates strong in-context learning capabilities. * The instruct fine-tuned model, `blip3-phi3-mini-instruct-r-v1`, achieves state-of-the-art performance among open-source and closed-source VLMs under 5b parameters. * `blip3-phi3-mini-instruct-r-v1` supports flexible high-resolution image encoding with efficient visual token sampling. More technical details will come with a technical report soon. # Datasets | Dataset Type| Dataset(s) Used | |--------|------------------------------------------| | Pretrain | caption data: (datacomp, cc12m, cc3m, SBU, vg) && interleaved data: obelics | | Instruction Tuning | LLaVA-Instruct-150K, ShareGPT4V captions, a mixture of academic VQA data including OCR/Document/Chart-focused tasks, publicly available text-only instruction data | # Results ### Pretrain | Model | Shot | COCO (val) | NoCaps (val) | TextCaps (val) | OKVQA (val) | TextVQA (val) | VizWiz (testdev) | VQAv2 (testdev) | |-------------|------|------------|--------------|----------------|--------------|---------------|------------------|-----------------| | Flamingo-3B | 4 | 85.0 | - | - | 43.3 | 32.7 | 34 | 53.2 | | | 8 | 90.6 | - | - | 44.6 | 32.4 | 38.4 | 55.4 | | MM1-3B | 0 | 73.5 | 55.6 | 63.3 | 26.1 | 29.4 | 15.6 | 46.2 | | | 4 | 112.3 | 99.7 | 84.1 | 48.6 | 45.3 | 38.0 | 57.9 | | | 8 | 114.6 | 104.7 | 88.8 | 48.4 | 44.6 | 46.4 | 63.6 | | **blip3-phi3-mini-base-r-v1 (Ours)**| 0 | **81.7** | **80.2** | 60.7 | **26.5** | **36.0** | **21.2** | **48.1** | | | 4 | 110.5 | **101.7** | **84.6** | **49.2** | **46.1** | **38.4** | **63.9** | | | 8 | 112.1 | 104.4 | 87.7 | **49.1** | **46.4** | 44.3 | **63.8** | ### Instruct | Model | SEED-IMG | MMBench(dev) | MME-total | MME-P | MME-C | MMStar | MMMU (val) | MMVet | MathVista (mini) | ScienceQA (test) | POPE | AI2D | | |----------------------------|----------|--------------|-----------|----------|---------|----------|------------|----------|------------------|------------------|----------|----------|---| | MM1-3B-Chat | 68.8 | 75.9 | 1761 | **1482** | 279 | - | 33.9 | 43.7 | - | - | **87.4** | - | | | openbmb/MiniCPM-V-2 | 67.1 | 69.6 | 1808 | - | - | - | 38.2 | - | 38.7 | - | - | - | | | VILA1.5-3B | 67.9 | 63.4 | - | 1442 | - | - | 33.3 | 35.4 | - | 69.0 | 85.9 | - | | | xtuner/llava-phi-3-mini-hf | 70.0 | 69.2 | 1790 | 1477 | 313 | 43.7 | **41.4** | - | - | 73.7 | 87.3 | 69.3 | | | **blip3-phi3-mini-instruct-r-v1 (Ours)** | **72.1** | **74.1** | **1827** | 1467 | **360** | **44.6** | 39.8 | **45.1** | **39.3** | **74.2** | 87.2 | **75.8** | | # Bias, Risks, Limitations, and Ethical Considerations We removed Laion from our training data due to known CSAM concerns. The other main data sources are from the internet, including webpages, image stock sites, and curated datasets released by the research community. 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 conduct an assessment of safety and fairness before applying to downstream applications. # How to use > We require the use of the development version (`"4.41.0.dev0"`) of the `transformers` library. To get it, as of 05/07/2024, one can use `pip uninstall -y transformers && pip install git+https://github.com/huggingface/transformers.` ```python from transformers import AutoModelForVision2Seq, AutoTokenizer, AutoImageProcessor, StoppingCriteria import torch import requests from PIL import Image # define the prompt template def apply_prompt_template(prompt): s = ( '<|system|>\nA chat between a curious user and an artificial intelligence assistant. ' "The assistant gives helpful, detailed, and polite answers to the user's questions.<|end|>\n" f'<|user|>\n\n{prompt}<|end|>\n<|assistant|>\n' ) return s class EosListStoppingCriteria(StoppingCriteria): def __init__(self, eos_sequence = [32007]): self.eos_sequence = eos_sequence def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor, **kwargs) -> bool: last_ids = input_ids[:,-len(self.eos_sequence):].tolist() return self.eos_sequence in last_ids # load models model_name_or_path = "Salesforce/blip3-phi3-mini-instruct-r-v1" model = AutoModelForVision2Seq.from_pretrained(model_name_or_path, trust_remote_code=True) tokenizer = AutoTokenizer.from_pretrained(model_name_or_path, trust_remote_code=True, use_fast=False, legacy=False) image_processor = AutoImageProcessor.from_pretrained(model_name_or_path, trust_remote_code=True) tokenizer = model.update_special_tokens(tokenizer) # craft a test sample img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg' raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB') query = "how many dogs are in the picture?" model = model.cuda() inputs = image_processor([raw_image], return_tensors="pt", image_aspect_ratio='anyres') prompt = apply_prompt_template(query) language_inputs = tokenizer([prompt], return_tensors="pt") inputs.update(language_inputs) inputs = {name: tensor.cuda() for name, tensor in inputs.items()} generated_text = model.generate(**inputs, image_size=[raw_image.size], pad_token_id=tokenizer.pad_token_id, do_sample=False, max_new_tokens=768, top_p=None, num_beams=1, stopping_criteria = [EosListStoppingCriteria()], ) prediction = tokenizer.decode(generated_text[0], skip_special_tokens=True).split("<|end|>")[0] print("==> prediction: ", prediction) # output: ==> prediction: There is one dog in the picture. ``` More comprehensive examples can be found in the [notebook](demo.ipynb). # Reproducibility: Our SFT evaluation is based on the VLMEvalKit, in which we fixed some inconsistencies with the official benchmarks (e.g., LLM judge API). During our development, we noticed that the raw resolution of the input image would noticeably affect the model output in some cases. # License Our code and weights are released under the Creative Commons Attribution Non Commercial 4.0 [LICENSE](LICENSE.txt). Please fill out a form at [here](https://forms.gle/ffPc9oZC2ZGeJ1N68) to consult the commercial use of model weights. # Code acknowledgement [LAVIS](https://github.com/salesforce/LAVIS) \ [openflamingo](https://github.com/mlfoundations/open_flamingo) \ [VLMEvalKit](https://github.com/open-compass/VLMEvalKit/tree/main) # Citation ``` @misc{blip3_phi3_mini, title={BLIP3-phi3-mini-instruct Model Card}, url={https://huggingface.co/Salesforce/blip3-phi3-mini-instruct-r-v1}, author={Salesforce AI Research}, month={May}, year={2024} } ``` # Troubleshoot 1. 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 ```