AIdeveric/llava-finetune-backup
📦 Companion evaluation repository for the LoRA-fine-tuned LLaVA-1.5-7B model for Structural Health Monitoring (SHM) of pavement distress.
📋 Overview
This repository contains the evaluation artifacts, training scripts, hyperparameter logs, and dataset files.
It complements the main model repository AIdeveric/llava-pavement-distress-lora by providing:
- ✅ Full reproducibility of the 74.31% Entity F1 result
- ✅ Complete Optuna hyperparameter search logs (30 trials)
- ✅ Baseline, LoRA, Unfreeze VLM, and ICL evaluation outputs
- ✅ Enhanced CFD dataset (151 images + GPT-4o descriptions + manual audit)
- ✅ Training/evaluation scripts for LLaMA-Factory integration
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support