[!IMPORTANT] This version of ColPali should be loaded with the transformers 🤗 release or Sentence Transformers, not with colpali-engine. It was converted using the convert_colpali_weights_to_hf.py script from the vidore/colpali-v1.3-merged checkpoint.

ColPali: Visual Retriever based on PaliGemma-3B with ColBERT strategy

ColPali is a model based on a novel model architecture and training strategy based on Vision Language Models (VLMs) to efficiently index documents from their visual features. It is a PaliGemma-3B extension that generates ColBERT- style multi-vector representations of text and images. It was introduced in the paper ColPali: Efficient Document Retrieval with Vision Language Models and first released in this repository

The HuggingFace transformers 🤗 implementation was contributed by Tony Wu (@tonywu71) and Yoni Gozlan (@yonigozlan).

Model Description

Read the transformers 🤗 model card: https://huggingface.co/docs/transformers/en/model_doc/colpali.

Model Training

Dataset

Our training dataset of 127,460 query-page pairs is comprised of train sets of openly available academic datasets (63%) and a synthetic dataset made up of pages from web-crawled PDF documents and augmented with VLM-generated (Claude-3 Sonnet) pseudo-questions (37%). Our training set is fully English by design, enabling us to study zero-shot generalization to non-English languages. We explicitly verify no multi-page PDF document is used both ViDoRe and in the train set to prevent evaluation contamination. A validation set is created with 2% of the samples to tune hyperparameters.

Note: Multilingual data is present in the pretraining corpus of the language model (Gemma-2B) and potentially occurs during PaliGemma-3B's multimodal training.

Parameters

All models are trained for 1 epoch on the train set. Unless specified otherwise, we train models in bfloat16 format, use low-rank adapters (LoRA) with alpha=32 and r=32 on the transformer layers from the language model, as well as the final randomly initialized projection layer, and use a paged_adamw_8bit optimizer. We train on an 8 GPU setup with data parallelism, a learning rate of 5e-5 with linear decay with 2.5% warmup steps, and a batch size of 32.

Usage

[!NOTE] This checkpoint was trained with the query prefix "Query: ", but the repository ships no processor_config.json, so transformers.ColPaliProcessor falls back to its class default of "Question: ". Set processor.query_prefix = "Query: " after loading the processor to reproduce the trained format. colpali-engine no longer sends this prefix either, as of 0.3.13 (illuin-tech/colpali#339). The Sentence Transformers configuration below restores it automatically.

Using Sentence Transformers

ColPali can be used as a multi-vector (ColBERT-style late interaction) retriever with Sentence Transformers via the MultiVectorEncoder:

pip install "sentence-transformers[image]>=6.0.0"

from sentence_transformers import MultiVectorEncoder

model = MultiVectorEncoder("vidore/colpali-v1.3-hf")

queries = [
    "What is the variable represented on the y-axis of the graph?",
    "Total outlay is maximum in which year?",
]
documents = [
    f"https://huggingface.co/datasets/sentence-transformers/example-documents/resolve/main/doc{i}.jpg" for i in range(1, 5)
]

query_embeddings = model.encode_query(queries)
document_embeddings = model.encode_document(documents)
print(query_embeddings[0].shape, document_embeddings[0].shape)
# (28, 128) (1030, 128)

similarities = model.similarity(query_embeddings, document_embeddings)
print(similarities)
# tensor([[22.3359, 19.8555, 19.6582, 19.0928],
#         [ 5.8828, 13.3398,  6.1621,  6.8135]])

Using transformers

import torch
from PIL import Image

from transformers import ColPaliForRetrieval, ColPaliProcessor

model_name = "vidore/colpali-v1.3-hf"

model = ColPaliForRetrieval.from_pretrained(
    model_name,
    torch_dtype=torch.bfloat16,
    device_map="cuda:0",  # or "mps" if on Apple Silicon
).eval()

processor = ColPaliProcessor.from_pretrained(model_name)
processor.query_prefix = "Query: "  # the prefix this checkpoint was trained with, see the note above

# Your inputs
images = [
    Image.new("RGB", (32, 32), color="white"),
    Image.new("RGB", (16, 16), color="black"),
]
queries = [
    "What is the organizational structure for our R&D department?",
    "Can you provide a breakdown of last year’s financial performance?",
]

# Process the inputs
batch_images = processor(images=images).to(model.device)
batch_queries = processor(text=queries).to(model.device)

# Forward pass
with torch.no_grad():
    image_embeddings = model(**batch_images)
    query_embeddings = model(**batch_queries)

# Score the queries against the images
scores = processor.score_retrieval(query_embeddings.embeddings, image_embeddings.embeddings)

Resources

• The ColPali arXiv paper can be found here. 📄
• The official blog post detailing ColPali can be found here. 📝
• The original model implementation code for the ColPali model and for the colpali-engine package can be found here. 🌎
• Cookbooks for learning to use the transformers-native version of ColPali, fine-tuning, and similarity maps generation can be found here. 📚

Limitations

• Focus: The model primarily focuses on PDF-type documents and high-ressources languages, potentially limiting its generalization to other document types or less represented languages.
• Support: The model relies on multi-vector retreiving derived from the ColBERT late interaction mechanism, which may require engineering efforts to adapt to widely used vector retrieval frameworks that lack native multi-vector support.

License

ColPali's vision language backbone model (PaliGemma) is under gemma license as specified in its model card. ColPali inherits from this gemma license.

Contact

• Manuel Faysse: manuel.faysse@illuin.tech
• Hugues Sibille: hugues.sibille@illuin.tech
• Tony Wu: tony.wu@illuin.tech

Citation

If you use any datasets or models from this organization in your research, please cite the original dataset as follows:

@misc{faysse2024colpaliefficientdocumentretrieval,
  title={ColPali: Efficient Document Retrieval with Vision Language Models}, 
  author={Manuel Faysse and Hugues Sibille and Tony Wu and Bilel Omrani and Gautier Viaud and Céline Hudelot and Pierre Colombo},
  year={2024},
  eprint={2407.01449},
  archivePrefix={arXiv},
  primaryClass={cs.IR},
  url={https://arxiv.org/abs/2407.01449}, 
}