st polish paraphrase from distilroberta

Introduction

The ST-Polish-Paraphrase-from-DistilRoBERTa is a sentence-transformers model designed for mapping sentences and paragraphs into a 768-dimensional dense vector space. This model is suitable for tasks such as clustering and semantic search, particularly for Polish language input.

Architecture

The model uses a SentenceTransformer architecture, which consists of:

• A Transformer component based on the RoBERTa model with a maximum sequence length of 256 tokens.
• A Pooling layer that averages the token embeddings to obtain sentence embeddings.

Training

The model was fine-tuned on paraphrasing tasks to enhance sentence similarity performance. It leverages the DistilRoBERTa backbone for efficient processing while maintaining performance.

Guide: Running Locally

1. Install Required Libraries

   • Use pip install -U sentence-transformers for the sentence-transformers library.

2. Using Sentence-Transformers

   from sentence_transformers import SentenceTransformer
   sentences = ["This is an example sentence", "Each sentence is converted"]
   model = SentenceTransformer('sdadas/st-polish-paraphrase-from-distilroberta')
   embeddings = model.encode(sentences)
   print(embeddings)
   

3. Using Hugging Face Transformers

   from transformers import AutoTokenizer, AutoModel
   import torch
   
   def mean_pooling(model_output, attention_mask):
       token_embeddings = model_output[0]
       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 = ['This is an example sentence', 'Each sentence is converted']
   tokenizer = AutoTokenizer.from_pretrained('sdadas/st-polish-paraphrase-from-distilroberta')
   model = AutoModel.from_pretrained('sdadas/st-polish-paraphrase-from-distilroberta')
   
   encoded_input = tokenizer(sentences, padding=True, truncation=True, return_tensors='pt')
   with torch.no_grad():
       model_output = model(**encoded_input)
   
   sentence_embeddings = mean_pooling(model_output, encoded_input['attention_mask'])
   print("Sentence embeddings:")
   print(sentence_embeddings)
   

4. Suggest Cloud GPUs

   • Consider using cloud services like AWS EC2, Google Cloud, or Azure for GPU resources to efficiently run the model, especially for large datasets or high-volume processing.

License

The model is licensed under the LGPL (Lesser General Public License), allowing for flexibility in use and modification while ensuring that derivative works remain open-source under the same license.