Starlings.Models.QwenChat
0.5.0
dotnet add package Starlings.Models.QwenChat --version 0.5.0
NuGet\Install-Package Starlings.Models.QwenChat -Version 0.5.0
<PackageReference Include="Starlings.Models.QwenChat" Version="0.5.0" />
<PackageVersion Include="Starlings.Models.QwenChat" Version="0.5.0" />
<PackageReference Include="Starlings.Models.QwenChat" />
paket add Starlings.Models.QwenChat --version 0.5.0
#r "nuget: Starlings.Models.QwenChat, 0.5.0"
#:package Starlings.Models.QwenChat@0.5.0
#addin nuget:?package=Starlings.Models.QwenChat&version=0.5.0
#tool nuget:?package=Starlings.Models.QwenChat&version=0.5.0
Starlings.Models.QwenChat
The model definitions of the experimental Qwen text chat profiles, without a runtime: Qwen/Qwen3-0.6B at c1899de289a04d12100db370d81485cdf75e47ca (rendered with enable_thinking=false) and Qwen/Qwen2.5-0.5B-Instruct at 7ae557604adf67be50417f59c2c2f167def9a775. No ONNX Runtime dependency (only Starlings.Generation.Abstractions, Starlings.Acquisition for the checkpoint locks, and the owned tokenizer Starlings.Tokenizers). To run the models on ONNX Runtime GenAI, reference Starlings.Models.QwenChat.OnnxGenAI, which brings this package. Experimental; the API may still change.
QwenChatTemplate.Render(checkpoint, messages, tools): the pinned templates rendered in C# for system, user, assistant and tool messages, tool definitions and assistant tool calls (tool calling experimental; multimodal input rejected). AQwenChatProfileof the runtime package converts implicitly to its checkpoint, soRender(profile, ...)works as before.QwenChatTemplate.For(checkpoint)(orcheckpoint.ChatTemplate) is the same renderer behindIChatTemplate.QwenChatToolCallParser: the templates' JSON tool-call form, whole or streamed; it also implementsIToolCallParser.QwenChatMessage,QwenChatTool,QwenChatToolCall,QwenChatOptions(withToGenerationOptions()),QwenChatPrompt; conversions to and fromChatTurn,ToolDefinitionandToolCall.QwenChatCheckpoint.Qwen3andQwenChatCheckpoint.Qwen25: repository, revision, model type, tokenizer and template SHA-256, end-of-sequence ids, decoder shape from the pinnedconfig.json,Profile(aModelProfile), andOpenTokenizer(path), which verifies the originaltokenizer.jsonbefore loading it.QwenChatCheckpoint.Lock: the asset lock of each Hub checkpoint itself (config.json,generation_config.json,model.safetensors,tokenizer.json,tokenizer_config.json,vocab.json,merges.txt; about 1.5 GB for Qwen3, 1.0 GB for Qwen2.5), for runtimes that load the original weights. The package also carries them aslocks/qwen3-0.6b.checkpoint.lock.jsonandlocks/qwen2.5-0.5b-instruct.checkpoint.lock.json(LockFileName; JSON lock schema version 2, read byModelAssetLock.Load). Nothing downloads unless you pass a lock toModelAssetStore.DownloadAsync. The locks leave out the repository'sREADME.mdandLICENSE(both checkpoints are Apache-2.0): a directory filled from a lock holds the weights without the license text, so fetchLICENSEfrom the Hub repository at the same revision if you redistribute them.
using Starlings.Models.QwenChat;
QwenChatCheckpoint checkpoint = QwenChatCheckpoint.Qwen3;
var tokenizer = checkpoint.OpenTokenizer(tokenizerPath); // digest and special ids checked
string text = QwenChatTemplate.Render(checkpoint, [new(QwenChatRole.User, "Explain gravity briefly.")]);
int[] promptIds = [.. tokenizer.Encode(text, addSpecialTokens: false).Ids];
Maturity
Maturity is per profile, generated from the registry manifests by model-pipeline catalog (do not edit this table by hand). Gates, evidence and capabilities per profile: the model catalog.
| Profile | Maturity | Platforms |
|---|---|---|
qwen2.5-0.5b-instruct/ortgenai-fp32-cpu |
experimental | linux-x64 cpu: experimental; windows-x64 cpu: untested |
qwen3-0.6b/ortgenai-fp32-cpu |
experimental | linux-x64 cpu: experimental; windows-x64 cpu: untested |
Attribution and license
The C# renderers adapt the chat_template strings of the pinned originals, Copyright 2024 Alibaba Cloud, Apache-2.0: see THIRD-PARTY-NOTICES.md and LICENSE.Qwen.txt in this package. This package's code is MIT OR Apache-2.0.
| 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
- Starlings.Acquisition (= 0.5.0)
- Starlings.Generation.Abstractions (= 0.5.0)
- Starlings.Tokenizers (= 0.3.4)
NuGet packages (1)
Showing the top 1 NuGet packages that depend on Starlings.Models.QwenChat:
| Package | Downloads |
|---|---|
|
Starlings.Models.QwenChat.OnnxGenAI
Qwen/Qwen3-0.6B (revision c1899de2, thinking disabled) and Qwen/Qwen2.5-0.5B-Instruct (revision 7ae55760) text chat generation in-process on ONNX Runtime GenAI 0.17.1: QwenChatModel over the model definitions Starlings.Models.QwenChat (the pinned chat templates rendered in C#, the tool-call parser, prompt ids from the owned tokenizer Starlings.Tokenizers: no Microsoft.ML.Tokenizers, no GenAI tokenizer) and Starling's decode loop (greedy, stop tokens and strings, streaming, cancellation between steps, usage). Ships no weights: QwenChatModelAssets downloads the profile's bundle explicitly from Starling's public Hugging Face hosting (or it is built locally from the pinned original), and Load verifies every file against the profile's digests. Profiles qwen3-0.6b/ortgenai-fp32-cpu and qwen2.5-0.5b-instruct/ortgenai-fp32-cpu, Linux x64. The application adds Microsoft.ML.OnnxRuntimeGenAI. Maturity is per model profile (Stable, Preview or Experimental): see the README, generated from the model registry. |
GitHub repositories
This package is not used by any popular GitHub repositories.