Records · kv:2610.00003v1

Reproduction of huggingface/pytorch-image-models: resnet26d on Imagenette

K-Veritas Team

Submitted by K-Veritas Team

Published 2026-10-07

Independent reproduction · Artifact evaluation · Computer vision

PaperReproduction of: original code

Abstract

Trained timm resnet26d from scratch on Imagenette (10-class ImageNet subset, 160 px release) with train.py: 40 epochs, image size 160, batch 128, native AMP, timm default SGD with cosine schedule and 5 warmup epochs, seed 42, 10-class head. The listed hfds/frgfm/imagenette dataset is a script-based Hugging Face dataset that current datasets releases cannot load, so the same Imagenette data was taken from the official fast.ai archive (imagenette2-160.tgz) as an image folder. Validation top-1/top-5 every epoch. Single NVIDIA RTX 5060 Ti 16 GB, PyTorch 2.11 cu128, Python 3.12.

Sealed report

Cite this record

@misc{kveritas261000003v1,
  title={Reproduction of huggingface/pytorch-image-models: resnet26d on Imagenette},
  author={K-Veritas Team},
  year={2026},
  howpublished={K-Veritas Records},
  note={kv:2610.00003v1},
  url={https://kveritas.org/records/2610.00003v1},
}

References

1. [1] Mamadou K. Keita and Christopher Homan. Computer Science Conferences Should Require Nonrepudiable Experimental Results. NeurIPS 2026 Position Paper Track. arXiv:2605.08586.

Sealed with K-Veritas [1]. Listed, not endorsed. A record shows these results came from the sealed code, unchanged. Whether the work is sound is for the reader to judge.