All Mpnet Base V2

🚀 all-mpnet-base-v2

This is a sentence-transformers model that maps sentences and paragraphs to a 384-dimensional dense vector space. It can be used for tasks such as clustering or semantic search.

🚀 Quick Start

📦 Installation

If you have sentence-transformers installed, using this model is straightforward:

pip install -U sentence-transformers

💻 Usage Examples

🔍 Basic Usage (Sentence-Transformers)

from sentence_transformers import SentenceTransformer
sentences = ["This is an example sentence", "Each sentence is converted"]

model = SentenceTransformer('sentence-transformers/all-mpnet-base-v2')
embeddings = model.encode(sentences)
print(embeddings)

🔍 Advanced Usage (HuggingFace Transformers)

Without sentence-transformers, you can use the model as follows: First, pass your input through the transformer model, then apply the appropriate pooling operation on top of the contextualized word embeddings.

from transformers import AutoTokenizer, AutoModel
import torch
import torch.nn.functional as F

#Mean Pooling - Take attention mask into account for correct averaging
def mean_pooling(model_output, attention_mask):
    token_embeddings = model_output[0] #First element of model_output contains all token embeddings
    input_mask_expanded = attention_mask.unsqueeze(-1).expand(token_embeddings.size()).float()
    return torch.sum(token_embeddings * input_mask_expanded, 1) / torch.clamp(input_mask_expanded.sum(1), min=1e-9)


# Sentences we want sentence embeddings for
sentences = ['This is an example sentence', 'Each sentence is converted']

# Load model from HuggingFace Hub
tokenizer = AutoTokenizer.from_pretrained('sentence-transformers/all-mpnet-base-v2')
model = AutoModel.from_pretrained('sentence-transformers/all-mpnet-base-v2')

# Tokenize sentences
encoded_input = tokenizer(sentences, padding=True, truncation=True, return_tensors='pt')

# Compute token embeddings
with torch.no_grad():
    model_output = model(**encoded_input)

# Perform pooling
sentence_embeddings = mean_pooling(model_output, encoded_input['attention_mask'])

# Normalize embeddings
sentence_embeddings = F.normalize(sentence_embeddings, p=2, dim=1)

print("Sentence embeddings:")
print(sentence_embeddings)

📚 Documentation

🔎 Evaluation Results

For an automated evaluation of this model, see the Sentence Embeddings Benchmark: https://seb.sbert.net

📖 Background

The project aims to train sentence embedding models on very large sentence-level datasets using a self-supervised contrastive learning objective. We used the pretrained microsoft/mpnet-base model and fine-tuned it on a 1B sentence pairs dataset. We use a contrastive learning objective: given a sentence from the pair, the model should predict which out of a set of randomly sampled other sentences was actually paired with it in our dataset.

We developed this model during the Community week using JAX/Flax for NLP & CV, organized by Hugging Face. We developed this model as part of the project: Train the Best Sentence Embedding Model Ever with 1B Training Pairs. We benefited from an efficient hardware infrastructure to run the project: 7 TPUs v3-8, as well as intervention from Google's Flax, JAX, and Cloud team members regarding efficient deep learning frameworks.

🎯 Intended uses

Our model is intended to be used as a sentence and short paragraph encoder. Given an input text, it outputs a vector that captures the semantic information. The sentence vector can be used for information retrieval, clustering, or sentence similarity tasks.

By default, input text longer than 512 word pieces is truncated.

⚙️ Training procedure

📋 Pre-training

We use the pretrained microsoft/mpnet-base model. Please refer to the model card for more detailed information about the pre-training procedure.

📋 Fine-tuning

We fine-tune the model using a contrastive objective. Formally, we compute the cosine similarity from each possible sentence pair in the batch. We then apply the cross-entropy loss by comparing with the true pairs.

⚙️ Hyperparameters

We trained our model on a TPU v3-8. We trained the model for 100k steps using a batch size of 1024 (128 per TPU core). We used a learning rate warm-up of 500. The sequence length was limited to 128 tokens. We used the AdamW optimizer with a 2e-5 learning rate. The full training script is available in this repository: train_script.py.

📊 Training data

We use the concatenation of multiple datasets to fine-tune our model. The total number of sentence pairs is over 1 billion sentences. We sampled each dataset with a weighted probability, and the configuration is detailed in the data_config.json file.

📄 License

This project is licensed under the Apache-2.0 license.