datracka/finance-analyzer

Finance Analyzer

Side project to learn about ML / AI, Model Fine tuning, PyTorch, Tensor Flow and Hugging Face libraries

run the project

pipenv shell

to start the virtual environment

pipenv install *

Virtual environment is managed previouly by pyenv https://realpython.com/intro-to-pyenv/

Folder csv with dataset is ignored by now.

Train a model

Currently last version of code is stores in trainer-2

• python trainer to tune a pretrained model. Pretrained model is bert-base-multilingual-uncased
  • created model store in /model folder (gitignored)
• python predictor and python pipeline to test model works

Also you have trainer_static_padding.py that is incompleted and not working.

to install package

pipenv install

Extra Ollama: Create modelfile (not used in this demo)

ollama create finance_analyzer_llama2 -f ./finance_analyer.txt ollama run finance_analyzer_llama2

TODOs

• (Optional) DeepLearning in https://course.fast.ai/ or https://www.deeplearning.ai/
• Learn how to use using Hugging face for loading a ollama model and learn about GGUF format
• improve accuracy of the model
• deploy it
• Improve accuracy
  • Simplify the dataset.
  • https://huggingface.co/PlanTL-GOB-ES/roberta-base-bne