app.py

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

title = """
# 👋🏻Welcome to 🙋🏻‍♂️Tonic's 🐣e5-mistral🛌🏻Embeddings """
description = """
You can use this Space to test out the current model [intfloat/e5-mistral-7b-instruct](https://huggingface.co/intfloat/e5-mistral-7b-instruct). e5mistral has a larger context window, a different prompting/return mechanism and generally better results than other embedding models. 
You can also use 🐣e5-mistral🛌🏻 by cloning this space. 🧬🔬🔍 Simply click here: <a style="display:inline-block" href="https://huggingface.co/spaces/Tonic/e5?duplicate=true"><img src="https://img.shields.io/badge/-Duplicate%20Space-blue?labelColor=white&style=flat&logo=data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAABAAAAAQCAYAAAAf8/9hAAAAAXNSR0IArs4c6QAAAP5JREFUOE+lk7FqAkEURY+ltunEgFXS2sZGIbXfEPdLlnxJyDdYB62sbbUKpLbVNhyYFzbrrA74YJlh9r079973psed0cvUD4A+4HoCjsA85X0Dfn/RBLBgBDxnQPfAEJgBY+A9gALA4tcbamSzS4xq4FOQAJgCDwV2CPKV8tZAJcAjMMkUe1vX+U+SMhfAJEHasQIWmXNN3abzDwHUrgcRGmYcgKe0bxrblHEB4E/pndMazNpSZGcsZdBlYJcEL9Afo75molJyM2FxmPgmgPqlWNLGfwZGG6UiyEvLzHYDmoPkDDiNm9JR9uboiONcBXrpY1qmgs21x1QwyZcpvxt9NS09PlsPAAAAAElFTkSuQmCC&logoWidth=14" alt="Duplicate Space"></a></h3> 
Join us : 🌟TeamTonic🌟 is always making cool demos! Join our active builder's🛠️community on 👻Discord:  [![Let's build the future of AI together! 🚀🤖](https://discordapp.com/api/guilds/1109943800132010065/widget.png)](https://discord.gg/GWpVpekp) On 🤗Huggingface: [TeamTonic](https://huggingface.co/TeamTonic) & [MultiTransformer](https://huggingface.co/MultiTransformer) On 🌐Github: [Polytonic](https://github.com/tonic-ai) & contribute to 🌟 [Poly](https://github.com/tonic-ai/poly)
"""
# Define the function to pool the last token
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]

# Define the function to get detailed instruct
def get_detailed_instruct(task_description: str, query: str) -> str:
    return f'Instruct: {task_description}\nQuery: {query}'

# Load tokenizer and model
tokenizer = AutoTokenizer.from_pretrained('intfloat/e5-mistral-7b-instruct')
model = AutoModel.from_pretrained('intfloat/e5-mistral-7b-instruct')

@spaces.GPU
def compute_embeddings(*input_texts):
    # Check if GPU is available and use it; otherwise, use CPU
    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

    # Move model to the chosen device
    model.to(device)
    max_length = 4096
    task = 'Given a web search query, retrieve relevant passages that answer the query'

    # Prepare the input texts
    processed_texts = [get_detailed_instruct(task, text) for text in input_texts]

    # Tokenize the input texts
    batch_dict = tokenizer(processed_texts, max_length=max_length - 1, return_attention_mask=False, padding=False, truncation=True)
    batch_dict['input_ids'] = [input_ids + [tokenizer.eos_token_id] for input_ids in batch_dict['input_ids']]
    batch_dict = tokenizer.pad(batch_dict, padding=True, return_attention_mask=True, return_tensors='pt')

    # Get model outputs
    outputs = model(**batch_dict)
    embeddings = last_token_pool(outputs.last_hidden_state, batch_dict['attention_mask'])

    # Normalize embeddings
    embeddings = F.normalize(embeddings, p=2, dim=1)
    return embeddings.detach().cpu().numpy()

    
def app_interface():
    with gr.Blocks() as demo:
        gr.Markdown(title)
        gr.Markdown(description)
        
        # Input text boxes
        input_text_boxes = [gr.Textbox(label=f"Input Text {i+1}") for i in range(4)]
        
        # Button to compute embeddings
        compute_button = gr.Button("Compute Embeddings")
        
        # Output display
        output_display = gr.Dataframe(headers=["Embedding"], datatype=["numpy"])
        
        # Layout
        with gr.Row():
            with gr.Column():
                for text_box in input_text_boxes:
                    text_box.render()
            with gr.Column():
                compute_button.render()
                output_display.render()

        # Function call
        compute_button.click(
            fn=compute_embeddings,
            inputs=input_text_boxes,
            outputs=output_display
        )

    return demo

# Run the Gradio app
app_interface().launch()