How to Use the STSB-MPNet-Base-V2 Model with Sentence-Transformers

Mar 27, 2024 | Educational

The STSB-MPNet-Base-V2 model is an impressive tool designed for mapping sentences to a 768-dimensional dense vector space. This means it can help with tasks like semantic search or clustering. In this guide, we will walk you through how to use this model for your applications, whether you are using the sentence-transformers library or Hugging Face Transformers library. Let’s get started!

Step 1: Installation

Before diving into the code, it’s crucial that you have the sentence-transformers library installed. If you haven’t installed it yet, you can do so with the following command:

pip install -U sentence-transformers

Step 2: Using the Model with Sentence-Transformers

Once the library is installed, using the model is quite straightforward. Here’s how you can perform sentence embeddings with it:

from sentence_transformers import SentenceTransformer

# Sample sentences
sentences = ["This is an example sentence", "Each sentence is converted"]

# Load the model
model = SentenceTransformer('sentence-transformers/stsb-mpnet-base-v2')

# Calculate the embeddings
embeddings = model.encode(sentences)

# Print the embeddings
print(embeddings)

Step 3: Using the Model without Sentence-Transformers

If you don’t want to use the sentence-transformers library, you can still access the model through Hugging Face Transformers. Here is how to do it:

from transformers import AutoTokenizer, AutoModel
import torch

# Mean Pooling function
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)

# Sample sentences
sentences = ["This is an example sentence", "Each sentence is converted"]

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

# Print the embeddings
print("Sentence embeddings:")
print(sentence_embeddings)

Understanding the Code: An Analogy

Think of the model as a sophisticated chef in a kitchen. The sentences you provide are like ingredients that the chef uses to prepare a dish. The chef (the model) takes these ingredients (sentences) and processes them, creating a unique flavor profile (embeddings) that represents the essence of the inputs. If you’re using the sentence-transformers library, it’s like having a pre-made recipe that guides you precisely through the steps in an easy way. Using Hugging Face Transformers without the sentence-transformers is like experimenting in the kitchen with your variations and methods. Both approaches will yield delicious results if done correctly!

Troubleshooting Tips

If you encounter issues while implementing the STSB-MPNet-Base-V2 model, here are some troubleshooting steps:

Ensure that the sentence-transformers library is properly installed and up to date.
Check that you have the correct model name, especially if you’re using the Hugging Face Transformers library.
If you’re running into memory issues, consider reducing the batch size of your sentences.
Ensure you have all necessary dependencies installed for both the sentence-transformers and Hugging Face libraries.

For more insights, updates, or to collaborate on AI development projects, stay connected with fxis.ai.

Conclusion

Using the STSB-MPNet-Base-V2 model can significantly enhance your ability to work with sentence embeddings, enabling you to tap into powerful functionalities for clustering and semantic analysis. At fxis.ai, we believe that such advancements are crucial for the future of AI, as they enable more comprehensive and effective solutions. Our team is continually exploring new methodologies to push the envelope in artificial intelligence, ensuring that our clients benefit from the latest technological innovations.