multilingual-e5-small embeddings (384-d, 100 languages)

Browser-ready import artifacts for feature-extraction, 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/intfloat/multilingual-e5-small/tree/614241f622f53c4eeff9890bdc4f31cfecc418b3
Upstream SHA-256 / commit	614241f622f53c4eeff9890bdc4f31cfecc418b3
Recipe	recipes/multilingual-e5-small.yaml — sha256 8e92c256e71e81d80f4ee39864cf6c8b3b6e8e9584ef641161f239a29a5a50be
Toolchain	Python 3.12.13, torch 2.10.0, onnx 1.23.0, onnxruntime 1.30.0 on Darwin 25.6.0 arm64
Converted	2026-09-22T19:27:46+00:00

Files

file	class	size	SHA-256
1_Pooling/config.json	bundle	0.00 MB	987f7a67a38fa564c849bb5d277c52ab9088a84368fc0be31a354125aebb12a0
config.json	bundle	0.00 MB	69137736cab8b8903a07fe8afaafdda25aac55415a12a55d1bffa9f581abf959
onnx/model.onnx	registry	448.48 MB	ca456c06b3a9505ddfd9131408916dd79290368331e7d76bb621f1cba6bc8665
special_tokens_map.json	bundle	0.00 MB	d05497f1da52c5e09554c0cd874037a083e1dc1b9cfd48034d1c717f1afc07a7
tokenizer.json	bundle	16.29 MB	0b44a9d7b51c3c62626640cda0e2c2f70fdacdc25bbbd68038369d14ebdf4c39
tokenizer_config.json	bundle	0.00 MB	a1d6bc8734a6f635dc158508bef000f8e2e5a759c7d92f984b2c86e5ff53425b

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	input_ids[1, 8], attention_mask[1, 8], token_type_ids[1, 8]	last_hidden_state[1, 8, 384]	2.1

Use in the browser

import * as ort from "onnxruntime-web";
const session = await ort.InferenceSession.create("https://huggingface.co/skillsafe-ai/multilingual-e5-small/resolve/main/onnx/model.onnx", { executionProviders: ["webgpu", "wasm"] });

Contract (onnx/model.onnx): input input_ids int64 ['batch_size', 'sequence_length'], attention_mask int64 ['batch_size', 'sequence_length'], token_type_ids int64 ['batch_size', 'sequence_length'] → output last_hidden_state float32 ['batch_size', 'sequence_length', 384]. Opset 11.

Licence and attribution

multilingual-e5-small: Liang Wang et al. (Microsoft), MIT License. https://huggingface.co/intfloat/multilingual-e5-small — the repo's own ONNX export.

Licence: MIT — notice: https://huggingface.co/intfloat/multilingual-e5-small/blob/614241f622f53c4eeff9890bdc4f31cfecc418b3/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.