Modelpublic

FasterAI-Labs/resnet50-imagenet1k-m31-fp32

sourceHugging Facebsd-3-clauseupdated 4d agoView on Hugging Face

0likes12downloads

Model Card

resnet50 — ImageNet-1k, -31.5 % MACs (Conv2d/Linear only), FP32

ImageNet-1k validation, 50000 images, one run (seed 42). Every published file is scored after reload, inside pinned numerics; the top-1 below is this artifact's own, not a comparison. The weights derive from the published checkpoint torchvision.models.resnet50 (IMAGENET1K_V2), which is the comparison point for size, memory and MACs only. The reload on a machine holding only the public package scores a frozen 2048-image subset of the same split.

Criteria

Top-1 on the evaluation set named above; size on disk, peak live activations and multiply-accumulates — the last two for one image of (3, 224, 224) at batch 1 — each against torchvision.models.resnet50 (IMAGENET1K_V2).

pruned FP32 — load with fastermodels.load(repo) (`model.safetensors`)

Top-1: 79.3 % [78.9, 79.6] on 50000 images

criterion	reference	this artifact	gap
size	102.5 MB, 25.6 M params	74.0 MB, 18.4 M params	-27.8 %
memory	9.6 MB	8.1 MB	-15.6 %
MACs	4.09 G	2.80 G	-31.5 %

Top-1 brackets give the 95 % interval over the evaluation images; size, memory and MACs gaps are against the reference.

Variants

Other points on the same ladder, from the same source model:

variant	repo	top-1, worst published form	size	memory	MACs
-31.5 % MACs (Conv2d/Linear only), INT8	`FasterAI-Labs/resnet50-imagenet1k-m31-int8`	78.6 %	18.8 MB	2.0 MB	2.80 G
-53.1 % MACs (Conv2d/Linear only), FP32	`FasterAI-Labs/resnet50-imagenet1k-m53-fp32`	77.4 %	50.7 MB	6.6 MB	1.92 G
-53.1 % MACs (Conv2d/Linear only), INT8	`FasterAI-Labs/resnet50-imagenet1k-m53-int8`	76.7 %	13.0 MB	1.7 MB	1.92 G
-66.2 % MACs (Conv2d/Linear only), FP32	`FasterAI-Labs/resnet50-imagenet1k-m66-fp32`	75.5 %	38.1 MB	5.7 MB	1.38 G
-66.2 % MACs (Conv2d/Linear only), INT8	`FasterAI-Labs/resnet50-imagenet1k-m66-int8`	74.6 %	9.8 MB	1.4 MB	1.38 G

Latency

not measured

Provenance

—Source model: torchvision.models.resnet50 (IMAGENET1K_V2)
—fasterai: 0.4.1
—fastermodels: 0.1.0
—torch: 2.9.1+cu128
—measured_on: 2026-09-21
—sourcestatehash: cf450479235c