LongShort-Llama-2-13B

Model Description

LongShort-Llama-2-13B is a large language model fine-tuned on earnings call documents to extract financial KPIs from the earnings call documents. It is based on the Llama-2-7B Architecture.

Model creator: Brief AI
Original model: Llama 2 13B Chat

Dataset Description

Data Source: Factiva
Data Description: 28K+ Earnings Call Documents
Data Scope: 1K+ public companies
Fine Tuning Data: Collection of 60K+ samples.

Prompt template: LongShort-Llama-2-13B

[INST]Given the context, answer the question.

### Question:
Extract all the finance-based performance indicators and evaluation metrics.

### Context:
{context}

### Answer:
[/INST]

Basics

This section provides information about the model type, version, license, funders, release date, developers, and contact information. It is useful for anyone who wants to reference the model.

Developed by: Brief AI Team

Model Type: Transformer-based Large Language Model

Version: 1.0.0

Languages: English

License: Apache 2.0

Release Date Estimate: Wednesday, 29.November.2023

Send Questions to: vishalparameswaran96@gmail.com

Cite as: Brief AI LongShort Language Model

Funded by: UChicago Data Science Institute

Mentored by: Nick Kadochnikov

Technical Specifications

This section includes details about the model objective and architecture, and the compute infrastructure. It is useful for people interested in model development.

Please see the LongShort training README for full details on replicating training.

Model Architecture and Objective

Modified from Llama-2-13B

Objective: Financial KPI extraction from earnings call documents.

Hardware and Software - Compute Infrastructure

4 NVIDIA L4 GPUs & 48 vCPUs
Environment: PyTorch (pytorch-2.0 w/ CUDA-11.8; see Github link)
CPU: GCP G2 Standard 48 (Platform: Intel Cascade Lake) (Accelerator Optimized)
CPU memory: 192GB RAM
GPU memory: 30GB per GPU

Training

This section provides information about the training. It is useful for people who want to learn more about the model inputs and training footprint.

The following bits and bytes quantization config was used during training:

quant_method: bitsandbytes
load_in_8bit: False
load_in_4bit: True
llm_int8_threshold: 6.0
llm_int8_skip_modules: None
llm_int8_enable_fp32_cpu_offload: False
llm_int8_has_fp16_weight: False
bnb_4bit_quant_type: nf4
bnb_4bit_use_double_quant: True
bnb_4bit_compute_dtype: float16

Framework versions

PEFT 0.4.0

Training Data

This section provides a high-level overview of the training data. It is relevant for anyone who wants to know the basics of what the model is learning.

Details for the dataset can be found in LongShort Dataset

Training data includes:

5000 Earnings Call Documents

How to use

This model can be easily used and deployed using HuggingFace's ecosystem. This needs transformers and accelerate installed. The model can be downloaded as follows:

LongShort-Llama-2-13B

Intended Use

This model is being created in order to enable public research on large language models (LLMs). LLMs are intended to be used for language generation or as a pre-trained base model that can be further fine-tuned for specific tasks. The use cases below are not exhaustive.

Direct Use

Text generation
Exploring characteristics of language generated by a language model
- Examples: Cloze tests, counterfactuals, generations with reframings

Downstream Use

Tasks that leverage language models include: Information Extraction, Question Answering, Summarization

Out-of-scope Uses

Using the model in high-stakes settings is out of scope for this model. The model is not designed for critical decisions nor uses with any material consequences on an individual's livelihood or wellbeing. The model outputs content that appears factual but may not be correct.

Out-of-scope Uses Include:

Usage for evaluating or scoring individuals, such as for employment, education, or credit
Applying the model for critical automatic decisions, generating factual content, creating reliable summaries, or generating predictions that must be correct

Misuse

Intentionally using the model for harm, violating human rights, or other kinds of malicious activities, is a misuse of this model. This includes:

Spam generation
Disinformation and influence operations
Disparagement and defamation
Harassment and abuse
Deception
Unconsented impersonation and imitation
Unconsented surveillance
Generating content without attribution to the model, as specified in the RAIL License, Use Restrictions

Intended Users

Direct Users

General Public
Researchers
Students
Educators
Engineers/developers
Non-commercial entities
Financial Industry

Risks and Limitations

This section identifies foreseeable harms and misunderstandings.

Model may:

Overrepresent some viewpoints and underrepresent others
Contain stereotypes
Contain personal information
Generate:
- Hateful, abusive, or violent language
- Discriminatory or prejudicial language
- Content that may not be appropriate for all settings, including sexual content
Make errors, including producing incorrect information as if it were factual
Generate irrelevant or repetitive outputs
Induce users into attributing human traits to it, such as sentience or consciousness

Evaluation

This section describes the evaluation protocols and provides the results.

Result: LongShort-Llama-2-13B gives 43.4% accuracy on a validation set of 10% of the original training dataset.

Train-time Evaluation:

Final checkpoint after 700 epochs:

Training Loss: 1.187

Recommendations

This section provides information on warnings and potential mitigations.

Indirect users should be made aware when the content they're working with is created by the LLM.
Users should be aware of Risks and Limitations, and include an appropriate age disclaimer or blocking interface as necessary.
Users of the model should provide mechanisms for those affected to provide feedback, such as an email address for comments.

Model Card Authors

Vishal Parameshwaran, Garima Sohi, Jose Gerala, Sanchit Narayan Kumar