FinanceConnect 13B

Table of Contents

Model Overview

The FinanceConnect model is a state-of-the-art, open-source chat model designed specifically for finance and economic discussions. Built on the robust Llama2-13B architecture, this model has been fine-tuned on a combination of FinTalk-19k and Alpaca datasets, making it a valuable resource for finance professionals, researchers, and enthusiasts.

Tasks and rank

Text Generation 1129th

Sentiment Analysis 51st

Natural Language Understanding 857th

Named Entity Recognition (NER) 96th

Chatbot 746th

Learn More About the Ranking System

Capabilities

Primary Tasks

The FinanceConnect model excels in the following areas:

• Financial knowledge: It has been fine-tuned on a combination of FinTalk-19k and Alpaca datasets, making it a valuable resource for financial discussions.
• Content generation: It can generate high-quality text related to finance, including articles, reports, and more.
• Research assistance: It can help researchers and professionals find relevant information and insights on various financial topics.

Strengths

The FinanceConnect model has several strengths that make it stand out:

• Domain specialization: It’s specifically designed for finance conversations, making it a go-to resource for financial professionals and researchers.
• Efficient performance: It’s optimized for both CPU and GPU platforms, ensuring efficient performance and fast response times.
• Comprehensive finance dataset: It utilizes a combination of comprehensive finance datasets, enabling it to provide accurate and up-to-date information.

Alternatives

NeMo SFT with Llama 2 10500 Fine-tuned language model

Bert Base Uncased 1831 Uncased BERT model

Roberta Base 399 Masked language model

Xlm Roberta Large 323 Multilingual Language Model

Bert Base Multilingual Uncased 111 Multilingual text model

Performance

Speed

How fast can the model generate text? With its efficient performance across both CPU and GPU platforms, it can generate high-quality text at a rapid pace. For example, when generating text based on a prompt, it can produce output in just a few seconds.

Prompt	Time Taken to Generate Output
List in detail ten key factors influencing the current state of the global economy.	2.50 seconds
Explain the concept of quantitative easing and its impact on financial markets.	1.80 seconds
I’ve recently started working and want to manage my finances. Give me a detailed answer in 10 points on how do I set up a budget?	3.20 seconds

Accuracy

How accurate is the model in generating text? With its high accuracy in text classification tasks, it can produce text that is both informative and relevant to the prompt.

Efficiency

How efficient is the model in processing large-scale datasets? With its ability to process large datasets efficiently, it can generate high-quality text even when dealing with complex financial topics.

Benchmark	FinanceConnect 13B MMLU	==BloombergGPT 50B==
MMLU	52.08	39.8
FPB	57.2	51.1
Cost	$27	$2.67 Million

Example Use Cases

Examples

What are the key factors that influence the current state of the global economy? The current state of the global economy is influenced by several factors, including: Globalization, Technological advancements, Political factors, Demographics, Environmental factors, International trade, Financial factors, Infrastructure, Education, and Health.

Explain the concept of quantitative easing and its impact on financial markets. Quantitative easing is a monetary policy tool used by central banks to stimulate economic growth and combat inflation. It involves the purchase of assets such as government bonds, mortgage-backed securities, and corporate bonds by a central bank.

Provide a detailed answer in 10 points on how to set up a budget. To set up a budget, follow these steps: Determine your income and expenses, Create a budget that includes all necessary expenses, Set aside a portion of your income for savings and investments, Monitor your spending regularly and adjust your budget as needed, Use cash or debit cards instead of credit cards to avoid overspending, Cut back on unnecessary expenses, Consider using a budgeting app or tool to help track your spending, Prioritize your spending based on your financial goals, Review your budget regularly and make adjustments as needed, Be flexible and willing to make changes to your budget as your financial situation changes.

The model can be used in a variety of ways, such as:

• Generating financial reports: It can generate high-quality financial reports, including analysis and insights.
• Research assistance: It can help researchers and professionals find relevant information and insights on various financial topics.
• Financial content creation: It can generate high-quality financial content, including articles, blog posts, and more.

Limitations

The model is not perfect and has several limitations, including:

• Out-of-scope use: It’s specifically designed for financial discussions and knowledge, and may not perform well in other areas.
• Dataset biases: The FinTalk-19k and Alpaca datasets used to train the model may contain inherent biases that influence the model’s outputs.
• Over-reliance: The model is an aid, not a replacement for human expertise, and decisions should be made with careful consideration and not solely based on the model’s outputs.

Format

The model uses the Llama2-13B architecture and accepts input in the form of text sequences. The input text can be a question, a statement, or a prompt, and the model will generate a response based on its understanding of the input.

Prerequisites

To get started with the model, you’ll need to have the following packages installed:

• torch
• transformers
• typing

You can install these packages using pip:

pip install torch transformers typing

Input and Output

Here’s an example of how to use the model to generate a response:

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

# Load the model and tokenizer
model_name = "ceadar-ie/FinanceConnect-13B"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

# Define a function to generate text
def generate_text(input_text):
    # Tokenize the input text
    inputs = tokenizer(input_text, return_tensors="pt")

    # Generate a response
    outputs = model.generate(inputs["input_ids"], max_length=1000)

    # Convert the response to text
    response = tokenizer.decode(outputs[0], skip_special_tokens=True)

    return response

# Test the function
input_text = "What is the current state of the global economy?"
response = generate_text(input_text)
print(response)

This code loads the model and tokenizer, defines a function to generate text, and tests the function with a sample input. The response will be a text sequence that provides information on the current state of the global economy.