CoolFace
Modelpublic

DIA-MVP/tinyllama-lora-a100

sourceHugging Faceapache-2.0updated 3mo agoView on Hugging Face
0likes11downloads
Model Card

TinyLlama 1.1B Chat — LoRA (NVIDIA A100)

A demo model from the Data & Impact Accounting (DIA) lab. It performs instruction-tuning (LoRA adapter) via LoRA (PEFT), with the base model TinyLlama/TinyLlama-1.1B-Chat-v1.0, trained on NVIDIA A100.

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

MetricValue
Hardware1× NVIDIA A100-SXM4-80GB
Compute0.035 GPU-hours
Energy0.0124 (measured) kWh
Carbon0.0008 (measured) kgCO₂eq
Water0.022–0.05 (estimated-from-default-wue) L
Grid regionca-on

Energy and carbon are measured with [CodeCarbon](https://github.com/mlco2/codecarbon); water is estimated from a default water-usage-effectiveness range. Carbon uses the local grid's intensity (Ontario, ~0.03 kgCO₂eq/kWh).

Reproduce

bash
REPO=DIA-MVP/tinyllama-lora-a100 python scripts/train_llama_lora.py

Links