TriadParty/deepmoney-34b-200k-base
TriadParty/deepmoney-34b-200k-base is a 34 billion parameter base model, pre-trained by TriadParty, specifically designed for financial analysis. It utilizes full-parameter pre-training on Yi-34b with an extended 200k context length, enabling deep comprehension of complex financial reports. This model excels at qualitative and quantitative financial analysis, making it ideal for building automated investment analysis pipelines.
Loading preview...
Deepmoney: A Specialized Financial Base Model
TriadParty/deepmoney-34b-200k-base is a 34 billion parameter model developed by TriadParty, forming the "Greed" installment in their Seven Deadly Sins series. This model is distinctively pre-trained on the Yi-34b architecture with a significant 200k context window, crucial for processing extensive financial documents like in-depth research reports.
Key Capabilities
- Specialized Financial Knowledge: Trained on high-quality, often institution-exclusive, research reports from 2019-2023, along with professional financial textbooks, rather than general public knowledge.
- Qualitative and Quantitative Analysis: Designed to master a wide array of qualitative and quantitative financial methods, essential for investment judgment.
- Data Extraction: Incorporates advanced techniques, including multi-modal models like cog-agent and emu2, for extracting data from tables and graphs within research reports.
- Pipeline Foundation: Serves as the base model for a comprehensive financial analysis pipeline, intended to support automated information collection, target judgment, and detailed analysis.
Good For
- Automated Investment Analysis: Ideal for developers looking to build systems that can interpret real-time financial news and data to make investment judgments.
- In-depth Financial Research: Its long context window and specialized training make it suitable for understanding and synthesizing complex financial reports.
- Quantitative Method Design: Capable of designing specific quantitative methods and identifying necessary data for analyzing market impacts of events, as demonstrated in comparative evaluations against GPT-4.