Applied AI project for financial and regulatory text classification using Mistral-based language models and Temporal Convolutional Network components.
The project compares baseline LLM adaptation approaches with TCN-based variants under limited GPU resources.
- Financial news classification
- ETS / regulatory regime classification
- AG News benchmark testing
- Baseline vs TCN-based model comparison
- Mistral-based text classification
- Temporal Convolutional Networks
- QLoRA baseline comparison
- PyTorch training pipeline
- Classification metrics
- Low-VRAM experimentation
src/— training, evaluation, and data loading scriptsscripts/— run examplesdocs/— notes and experiment summaries
Datasets, local model files, checkpoints, and logs are not included.
The repository contains code only.
Cleaned public version of a local applied AI research project.