Views
No views yet
| Metric | Baseline (Zero-Shot) | MeterMind (Fine-Tuned) | Improvement |
|---|---|---|---|
| MAE | 2.82 | 0.60 | 79% better |
| RMSE | 4.35 | 1.10 | 75% better |
| Within ±1 | 53% | 87% | +34 points |
| Exact Match | 27% | 60% | +33 points |
q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj, fc1, fc21from transformers import AutoProcessor, AutoModelForVision2Seq
2from peft import PeftModel
3
4# Load base model
5base_model = AutoModelForVision2Seq.from_pretrained("unsloth/ERNIE-4.5-VL-28B-A3B-PT")
6processor = AutoProcessor.from_pretrained("unsloth/ERNIE-4.5-VL-28B-A3B-PT")
7
8# Load LoRA adapter
9model = PeftModel.from_pretrained(base_model, "luliuzee/metermind-ernie-gauge-lora")
10
11# Inference
12prompt = "What is the reading on this gauge? Reply with only the numeric value."
13# ... process image and generate1@misc{metermind2024,
2 title={MeterMind: Industrial Gauge Reading with Fine-Tuned ERNIE-4.5-VL},
3 author={luliuzee},
4 year={2024},
5 publisher={HuggingFace}
6}