A demo model from the Data & Impact Accounting (DIA) lab. It performs binary sentiment classification (SST-2) via full fine-tune, with the base model distilbert-base-uncased, trained on NVIDIA A40.

The point of this repo is not the model itself but its dia_report — a standardized record of the energy, carbon, and water used to train it, embedded in this card's metadata.

This footprint feeds the DIA dashboard, which rolls up a base model and all its derivatives to show the cumulative carbon, water, and energy cost of a model family.

Training footprint

Metric	Value
Hardware	1× NVIDIA A40
Compute	0.0233 GPU-hours
Energy	0.0061 (measured) kWh
Carbon	0.0004 (measured) kgCO₂eq
Water	0.011–0.025 (estimated-from-default-wue) L
Grid region	ca-on

Energy and carbon are measured with CodeCarbon; water is estimated from a default water-usage-effectiveness range. Carbon uses the local grid's intensity (Ontario, ~0.03 kgCO₂eq/kWh).

Reproduce

REPO=DIA-MVP/my-bert-sentiment-a40 python scripts/train_bert_demo.py

DIA-MVP/my-bert-sentiment-a40

DistilBERT — SST-2 sentiment (NVIDIA A40)

Training footprint

Reproduce

Links

DIA-MVP/
my-bert-sentiment-a40