app.py

from transformers import LongformerTokenizer, LongformerModel
import torch
import gradio as gr

# Load the pre-trained Longformer model and tokenizer
tokenizer = LongformerTokenizer.from_pretrained('allenai/longformer-base-4096')
model = LongformerModel.from_pretrained('allenai/longformer-base-4096')

def get_longformer_embeddings(sentences):
    # Tokenize input sentences
    inputs = tokenizer(sentences, return_tensors='pt', padding=True, truncation=True, max_length=2048)
    # Get embeddings
    with torch.no_grad():
        outputs = model(**inputs)
    embeddings = outputs.last_hidden_state.mean(dim=1)  # Mean pooling over the sequence
    return embeddings.numpy().tolist()

# Define the Gradio interface
interface = gr.Interface(
    fn=get_longformer_embeddings,  # Function to call
    inputs=gr.Textbox(lines=2, placeholder="Enter sentences here, one per line"),  # Input component
    outputs=gr.JSON(),  # Output component
    title="Sentence Embeddings with Longformer",  # Interface title
    description="Enter sentences to get their embeddings with Longformer (up to 2048 tokens)."  # Description
)

# Launch the interface
interface.launch()