Skip to content

Adityag009/FinChat-AI

Repository files navigation

FinChat-AI

FinChat-AI is an advanced AI-powered project designed to analyze financial markets using Natural Language Processing (NLP) techniques. It leverages cutting-edge transformer models, such as ChatGLM2-6B, fine-tuned with LoRA (Low-Rank Adaptation), to provide insights into financial data and sentiment.

Features

  • Sentiment Analysis: Accurately determines the sentiment (positive, neutral, negative) of financial news and market-related text.
  • Pattern Recognition: Identifies technical patterns like the Inverted Head and Shoulders from textual market descriptions.
  • Fine-Tuned LoRA Weights: The repository includes pre-trained LoRA weights for fine-tuning ChatGLM2-6B, enabling efficient deployment of the model with reduced computational overhead.
  • Customizable Prompts: Users can input financial scenarios or market updates, and the model generates insightful analyses based on pre-trained knowledge and fine-tuning.
  • Transformer-based Architecture: Uses Hugging Face’s transformers library for seamless interaction and deployment of models.

Repository Content

  • Finchat2.ipynb: The primary notebook demonstrating how to load the model, interact with it, and perform tasks such as sentiment analysis and pattern recognition.
  • saved_model.zip: Contains fine-tuned LoRA weights, allowing users to replicate and further fine-tune the project on their datasets.
  • README.md: This file provides an overview of the project, its features, and usage instructions.
  • Training Resources: Scripts for training with LoRA and integrating with Meta-LLaMA-2 architecture.

Model and Training Details

  • Base Model: ChatGLM2-6B
  • Fine-Tuning: Utilizes LoRA for efficient parameter updates during training.
  • Training Dataset: Focused on financial data, market updates, and technical analysis descriptions.
  • Frameworks:
    • Hugging Face Transformers
    • PEFT for LoRA
    • PyTorch for model deployment

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors