skillsafe-ai/vit-base-patch16-224

ViT-base/16 ImageNet-1k classification

Browser-ready import artifacts for image-classification, produced by SkillSafe's reproducible converter (models/ in skillsafe.ai) from a pinned upstream source. Every byte here is derivable from that source plus the recipe below; nothing was edited by hand.

Provenance

Upstream	https://huggingface.co/Xenova/vit-base-patch16-224/tree/66fef688e8dbe77dd9d5aa256353f9ad8b0ef799
Upstream SHA-256 / commit	66fef688e8dbe77dd9d5aa256353f9ad8b0ef799
Recipe	recipes/vit-base-patch16-224.yaml — sha256 25150f00fb18af528acf2f49617699b573d5fef4dc30d53576acbb8ec05689bc
Toolchain	Python 3.12.13, torch 2.10.0, onnx 1.23.0, onnxruntime 1.30.0 on Darwin 25.6.0 arm64
Converted	2026-09-22T21:52:59+00:00

Files

file	class	size	SHA-256
config.json	bundle	0.07 MB	ba68592930f3a1aa36c96630d374018d02c09fa59a7e38a3978613cadba02af4
onnx/model.onnx	registry (fp32)	330.48 MB	4bafe23c7e2650856449a792eafcc1d3bab4a2f41bcf58c9f3eac99d98719fcc
onnx/model_quantized.onnx	registry (q8)	84.17 MB	fe13a717a54c0f6ef1a966e0dd67a82e532e0898018fc50f7c37d437c225074a
preprocessor_config.json	bundle	0.00 MB	b09d2030f83f2a59d12c717d41a9135a3f0c1ba0a2a5df694dbc40f77735daed

registry files are parameter files served from models.skillsafe.ai once vetted; bundle files ship inside an app; registry-shared is a runtime library reused by every model of the same architecture.

Verification

Imported as published upstream (no conversion). Each file is pinned by SHA-256 to its source; every ONNX file passed onnx.checker and a CPU smoke run under onnxruntime with zero-filled inputs at the declared shapes:

file	inputs	outputs	ms
onnx/model.onnx	pixel_values[1, 3, 224, 224]	logits[1, 1000]	25.2
onnx/model_quantized.onnx	pixel_values[1, 3, 224, 224]	logits[1, 1000]	19.3

Use in the browser

import * as ort from "onnxruntime-web";
const session = await ort.InferenceSession.create("https://huggingface.co/skillsafe-ai/vit-base-patch16-224/resolve/main/onnx/model.onnx", { executionProviders: ["webgpu", "wasm"] });

Contract (onnx/model.onnx): input pixel_values float32 ['batch_size', 'num_channels', 'height', 'width'] → output logits float32 ['batch_size', 1000]. Opset 11.

Licence and attribution

ViT base patch16 224: Google, Apache License 2.0 (https://huggingface.co/google/vit-base-patch16-224); ONNX export by Xenova.

Licence: Apache-2.0 — notice: https://huggingface.co/google/vit-base-patch16-224/blob/main/README.md. The conversion recipe and this model card are part of the SkillSafe repository and carry its licence; the weights remain under the upstream licence above.

The full manifest.json in this repo records the recipe, sources, toolchain (including the uv.lock hash) and per-file verification numbers.