Side project to learn about ML / AI, Model Fine tuning, PyTorch, Tensor Flow and Hugging Face libraries
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.
Currently last version of code is stores in trainer-2
python trainer
to tune a pretrained model. Pretrained model isbert-base-multilingual-uncased
- created model store in
/model
folder (gitignored)
- created model store in
python predictor
andpython pipeline
to test model works
Also you have trainer_static_padding.py
that is incompleted and not working.
pipenv install
ollama create finance_analyzer_llama2 -f ./finance_analyer.txt ollama run finance_analyzer_llama2
- (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