RecursiveAI.OneNlp.Models.Multilingual.FrozenEncoder
1.0.0
Prefix Reserved
dotnet add package RecursiveAI.OneNlp.Models.Multilingual.FrozenEncoder --version 1.0.0
NuGet\Install-Package RecursiveAI.OneNlp.Models.Multilingual.FrozenEncoder -Version 1.0.0
<PackageReference Include="RecursiveAI.OneNlp.Models.Multilingual.FrozenEncoder" Version="1.0.0" />
<PackageVersion Include="RecursiveAI.OneNlp.Models.Multilingual.FrozenEncoder" Version="1.0.0" />
<PackageReference Include="RecursiveAI.OneNlp.Models.Multilingual.FrozenEncoder" />
paket add RecursiveAI.OneNlp.Models.Multilingual.FrozenEncoder --version 1.0.0
#r "nuget: RecursiveAI.OneNlp.Models.Multilingual.FrozenEncoder, 1.0.0"
#:package RecursiveAI.OneNlp.Models.Multilingual.FrozenEncoder@1.0.0
#addin nuget:?package=RecursiveAI.OneNlp.Models.Multilingual.FrozenEncoder&version=1.0.0
#tool nuget:?package=RecursiveAI.OneNlp.Models.Multilingual.FrozenEncoder&version=1.0.0
RecursiveAI.OneNLP.Models.Multilingual.FrozenEncoder
spaCy-class NLP for .NET. The frozen distilled multilingual encoder of RecursiveAI.OneNlp: a six-layer, width-384 student of mmBERT-base (113 languages, 35.5 M weights, 142 MB as one ONNX file) and the model's tokenizer.json, embedded in the package. It maps the tokens of a Doc to 384-wide vectors, for your own heads or for Train. No downloads, no native code, NativeAOT and trim clean. Unaffiliated with other projects named OneNLP.
using RecursiveAI.OneNlp;
using RecursiveAI.OneNlp.Models.Multilingual.FrozenEncoder;
using RecursiveAI.OneNlp.Neural.Encoders;
using TransformerEncoder encoder = FrozenStudentEncoder.Create(); // IEncoder, width 384
using Pipeline tokenizer = Pipeline.Blank(Language.German);
Doc doc = tokenizer.Process("Der Hund läuft schnell.");
EncodedBatch batch = encoder.Encode([doc]); // one 384-wide row per token
FrozenStudentEncoder.Create reads the embedded weights; keep one encoder and share it between threads. Use it wherever an IEncoder is taken, for example as the encoder of a head you train with Train. The raw files are available too: Part.Open() (the ONNX graph) and Part.OpenTokenizer().
The student was distilled from jhu-clsp/mmBERT-base (MIT) on FineWeb-2 (ODC-By 1.0, subject to Common Crawl's terms) and Wikipedia text (CC BY-SA 3.0); the weights therefore carry share-alike lineage, credited in THIRD-PARTY-NOTICES.md. PROVENANCE.md lists every source, recipe and seed. The weights are licensed MIT OR Apache-2.0.
- Source and documentation: https://github.com/recursive-ai-oss/onenlp-csharp
- Home: https://recursiveai.org/onenlp/
Copyright (c) Recursive AI, LLC.
| Product | Versions Compatible and additional computed target framework versions. |
|---|---|
| .NET | net10.0 is compatible. net10.0-android was computed. net10.0-browser was computed. net10.0-ios was computed. net10.0-maccatalyst was computed. net10.0-macos was computed. net10.0-tvos was computed. net10.0-windows was computed. |
-
net10.0
- RecursiveAI.OneNlp.Neural (>= 1.0.0)
NuGet packages (1)
Showing the top 1 NuGet packages that depend on RecursiveAI.OneNlp.Models.Multilingual.FrozenEncoder:
| Package | Downloads |
|---|---|
|
RecursiveAI.OneNlp.Models.Multilingual
spaCy-class NLP for .NET: the distilled accurate-tier model (a six-layer student of mmBERT-base, fine-tuned) for English, German, French, Spanish, Portuguese, Italian, Dutch, Turkish and Chinese, plus an open-type entity recognizer. Tags, morphology, lemmas, dependency parse and entities for the Nlp verbs; the Nlp verbs use it for these languages when no fast-tier package is referenced. About 500 MB, split over several packages. No downloads. Unaffiliated with other projects named OneNLP. |
GitHub repositories
This package is not used by any popular GitHub repositories.
| Version | Downloads | Last Updated |
|---|---|---|
| 1.0.0 | 119 | 11/8/2026 |
See CHANGELOG.md in the repository.