FinGLM is an open, public-finance oriented large model project. It provides project structure, data flows, and multiple open-source demos across teams, with a focus on finance data processing and Q&A workflows.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
FinGLM appears as an ecosystem and set of open-source components for financial large language model use. It aggregates multiple teams' open-source projects, tutorials, and data handling workflows related to processing annual reports and building finance QA capabilities.
How it works
The repo describes a framework for data preparation, model fine-tuning, and QA workflows: data extraction from PDFs to TXT/HTML, data segmentation, basic financial feature calculations, and constructing databases and vector stores. It includes prompts and example QA pipelines, and mentions integrating with vector stores and LangChain for applications.
Getting started
- The repository content is organized around project overviews, data workflows, and open-source showcases of participating teams. The README provides instructions and scripts for data processing steps and demonstrations, including data preparation, model fine-tuning, and QA flows. There is no explicit installation command block in the truncated README snippet provided.
Recent releases
- RELEASES (latest 0): - none
Traction
- Stars: 2255
- Forks: 315
- Open issues: 7
Behind the repo
- This is a collaborative open-source effort tied to Chinese fintech/AI challenge teams and a public data set including annual reports. The README lists multiple contributing teams and example projects under FinGLM_all, MantouTech, NanNaTeam, Chatglm, nsddd, Longying, and others.
Caveats
- Language: HTML (repository language tag is HTML)
- License: none listed
- Created: 2023-09-25
- Last push: 2024-05-08
- Data set references and data licenses are described; no explicit license is provided in the snippet.