app.py

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

def last_token_pool(last_hidden_states: Tensor,
                 attention_mask: Tensor) -> Tensor:
    left_padding = (attention_mask[:, -1].sum() == attention_mask.shape[0])
    if left_padding:
        return last_hidden_states[:, -1]
    else:
        sequence_lengths = attention_mask.sum(dim=1) - 1
        batch_size = last_hidden_states.shape[0]
        return last_hidden_states[torch.arange(batch_size, device=last_hidden_states.device), sequence_lengths]

def get_similarity_scores(queries:list, passages:list, model, tokenizer):
    print("queries", queries)
    print("passages", passages)
    tokenizer.add_eos_token = True

    max_length = 4096
    input_texts = queries + passages
    batch_dict = tokenizer(input_texts, max_length=max_length - 1, padding=True, truncation=True, return_tensors="pt")
    outputs = model(**batch_dict)
    embeddings = last_token_pool(outputs.last_hidden_state, batch_dict['attention_mask'])

    embeddings = F.normalize(embeddings, p=2, dim=1)
    scores = (embeddings[:len(queries)] @ embeddings[len(queries):].T) * 100
    return scores.tolist()

def similarity_ui(keyNames, fields):
    print("keynames", keyNames)
    print("fields", fields)
    task = 'Given a keyName, find similarity score against provided fields'
    queries = keyNames.split(',')
    passages = fields.split(',')

    scores = get_similarity_scores(queries, passages, model, tokenizer)
    return scores

# Load model and tokenizer
tokenizer = AutoTokenizer.from_pretrained('Salesforce/SFR-Embedding-Mistral')
model = AutoModel.from_pretrained('Salesforce/SFR-Embedding-Mistral')

# Create Gradio Interface
gr.Interface(
    fn=similarity_ui, 
    inputs=[gr.Textbox(), gr.Textbox()],
    outputs=gr.Textbox(),
    title="Similarity Score Calculator",
    description="Enter a Key Name and 3 Fields to find similarity scores"
).launch()