You can deploy this model and start earning money today!

Pre-built tokie tokenizer included (tokenizer.tkz). 5x faster tokenization, drop-in replacement for HuggingFace tokenizers.

potion-retrieval-32M Model Card

Name: potion-retrieval-32M
Availability: InStock

Model2Vec logo

This Model2Vec model is optmized for retrieval tasks. It is a finetune of potion-base-32M. It's finetuned using a modified version of the training approach described in this blogpost. It uses static embeddings, allowing text embeddings to be computed orders of magnitude faster on both GPU and CPU. It is designed for applications where computational resources are limited or where real-time performance is critical.

Installation

Install model2vec using pip:

pip install model2vec

Usage

Load this model using the from_pretrained method:

python

from model2vec import StaticModel
# Load a pretrained Model2Vec model
model = StaticModel.from_pretrained("minishlab/potion-retrieval-32M")
# Compute text embeddings
embeddings = model.encode(["Example sentence"])

How it works

Model2vec creates a small, static model that outperforms other static embedding models by a large margin on all tasks on MTEB. This model is pre-trained using Tokenlearn. It's created using the following steps:

Distillation: first, a model is distilled from a sentence transformer model using Model2Vec.
Training data creation: the sentence transformer model is used to create training data by creating mean output embeddings on a large corpus.
Training: the distilled model is trained on the training data using Tokenlearn.
Post-training re-regularization: after training, the model is re-regularized by weighting the tokens based on their frequency, applying PCA, and finally applying SIF weighting.

The results for this model can be found on the Model2Vec results page.

Results

Model	MTEB Retrieval Score
all-MiniLM-L6-v2	42.92
potion-retrieval-32M	35.06
static-retrieval-mrl-en-v1	34.95
potion-base-32M	32.67

The results show that potion-retrieval-32M is the most performant static retrieval model, reaching 81.69% of the performance of all-MiniLM-L6-v2 with a retrieval score of 35.06 while being orders of magnitude faster.

For full results, see the MTEB leaderboard.

Additional Resources

Library Authors

Model2Vec was developed by the Minish Lab team consisting of Stephan Tulkens and Thomas van Dongen.

Citation

Please cite the Model2Vec repository if you use this model in your work.

bibtex

@software{minishlab2024model2vec,
  author       = {Stephan Tulkens and {van Dongen}, Thomas},
  title        = {Model2Vec: Fast State-of-the-Art Static Embeddings},
  year         = {2024},
  publisher    = {Zenodo},
  doi          = {10.5281/zenodo.17270888},
  url          = {https://github.com/MinishLab/model2vec},
  license      = {MIT}
}

Reproducibility

The following script can be used to reproduce this model. All credits go to Tom Aarsen for this fine-tuning approach and code he introduced in his blogpost. We make a few modifcations to the original code, namely:

We start with a pre-trained Model2Vec model (potion-base-32M).
We reduce the dataset size by a factor of 10. During experiments we saw that we didn't need the full dataset for the model to converge.
We decease the learning rate and train for 3 epochs instead of 1. Using a high learning rate wipes the effects of using a pre-trained model.