Skip to content
View Abandon-ht's full-sized avatar

Highlights

  • Pro

Block or report Abandon-ht

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Abandon-ht/README.md

小佬鼠 / LittleMouse — 才疏学浅的小佬鼠。Curious about models, machines, music and minds.

项目 / Projects · 模型 / Models · 方向 / Focus · 关于 / About

🤗 Hugging Face / Abandon-ht · X / @Abandon_ht · Bilibili ↗

你好,我是 小佬鼠 · LittleMouse,关注端侧模型部署、语音交互与嵌入式系统。我的实践围绕 ASR、TTS、LLM 与计算机视觉展开,从模型导出与图适配、量化编译,到 C++ / Rust 运行时集成和真实设备上的交互应用。

Hi, I'm LittleMouse (小佬鼠). I focus on on-device model deployment, speech interaction and embedded systems, working across ASR, TTS, LLMs and computer vision—from model export, graph adaptation and quantization to C++ / Rust runtime integration and interactive applications on real devices.

C / C++ · Python · Rust · PyTorch · ONNX · Tauri · Linux

精选项目 · Featured Projects

DesktopPet:本地语音交互与 Live2D 桌面伙伴;A Live2D companion with local voice YOLO26 / NPU:模型导出、NPU 编译与推理;Model export, NPU compilation and inference
sherpa-onnx:AX650 ZipVoice TTS 后端贡献;AX650 ZipVoice TTS backend contribution m3axpi_model:YOLOv5 / YOLOv8 端侧部署示例;On-device YOLOv5 / YOLOv8 examples
Edge AI Guide:检测、分割与姿态模型部署;Detection, segmentation and pose deployment ModuleLLM Guide:ESP32 与 PC 端开发资料;ESP32 and PC development notes

工具与更多探索 · Tools & More

模型与导出资产 · Models & Export Artifacts

在 Hugging Face 分享语音、语言和视觉模型的转换产物与部署资料。
I share converted models, export artifacts and deployment notes for speech, language and vision on Hugging Face.

方向 · Area 公开仓库 · Public Repositories
语音识别 · ASR SenseVoiceSmall · Qwen3-ASR · Parakeet TDT
语音合成 · TTS Kokoro · Supertonic · Chatterbox
本地大模型 · LLM Qwen3-4B · Qwen2.5-0.5B Merge
视觉推理 · Vision YOLO11 · Palm Detection · Hand Pose

→ Hugging Face / Abandon-ht — 查看全部模型 · Explore all models

技术方向与工程实践 · Focus & Engineering

方向 · Area 实践内容 · Engineering Focus
ASR / TTS 多模型适配、VAD、文本前端与语音运行时集成。
Multi-model adaptation, VAD, text frontends and speech runtime integration.
LLM 本地推理、prefill / KV cache 工作流与分词器工具。
Local inference, prefill / KV-cache workflows and tokenizer tools.
计算机视觉 · Vision 检测、分割、姿态与多媒体推理链路。
Detection, segmentation, pose and multimedia inference pipelines.
部署工程 · Deployment ONNX 图适配、量化校准、NPU 编译与 CPU / NPU 协同。
ONNX graph adaptation, quantization, NPU compilation and CPU / NPU integration.
嵌入式交互 · Embedded ESP32 / Linux、音频采集、UART / I2S 与应用编排。
ESP32 / Linux, audio I/O, UART / I2S and application orchestration.

端侧语音交互 / On-device voice: Audio → VAD / KWS → ASR → LLM → TTS → Playback

开源参与 · Open Source

为 sherpa-onnx 提交 AX650 ZipVoice TTS 后端实现,让语音合成模型在端侧硬件上运行。
Contributed an AX650 ZipVoice TTS backend proposal to sherpa-onnx, bringing speech synthesis to on-device hardware.

→ 上游提案与实现 · Upstream proposal & implementation: PR #4038

欢迎在对应项目的 Issues / Discussions 中交流模型部署、设备适配与使用体验。
Feel free to discuss model deployment, device integration and feedback in the relevant project's Issues / Discussions.

关于小佬鼠 · About LittleMouse

才疏学浅的小佬鼠,还在学习,也还在拆解世界。
A curious LittleMouse, always learning—and taking the world apart to understand it.

我容易被「有结构、有原理、有思想,能拆开研究」的东西吸引。
I'm drawn to things with structure, underlying principles and ideas worth taking apart.

代码之外,我喜欢什么? · What catches my curiosity beyond code?
兴趣 · Interest 我喜欢拆解的东西 · What I Like to Explore
知识 · Ideas 哲学、认知科学、语言与思维,以及神学、历史和思想。意识如何形成?语言与抽象思考是什么关系?自由意志与决定论如何相处?
Philosophy, cognitive science, language and thought, theology and history—especially consciousness, abstract thinking, free will and determinism.
科技 · Systems Apple Silicon、CPU / GPU / NPU 架构、操作系统、统一内存、HDR、音频系统与性能分析。
Apple Silicon, CPU / GPU / NPU architectures, operating systems, unified memory, HDR, audio systems and performance analysis.
人工智能 · AI Transformer、Attention、KV cache、量化与推理加速;vLLM、SGLang、llama.cpp、MLX;ASR、TTS、Agent、多模态、论文与 benchmark。也喜欢自己跑模型、做测试。
Model architectures, inference systems, speech, agents and multimodal AI. I enjoy reading papers, running models and testing how they perform.
古典音乐 · Classical Music 尤其喜欢巴赫与复调,也喜欢肖邦、贝多芬、舒伯特、德彪西等。主题、对题、卡农、倒影,以及和声、织体与乐曲结构,都值得细听。
Especially Bach and counterpoint, alongside Chopin, Beethoven, Schubert, Debussy and others—subjects, countersubjects, canons, inversion, harmony, texture and musical form.
游戏 · Games 有世界观、复杂叙事、哲学问题或独特系统的作品;也会研究 NPC AI、图形渲染、超分辨率与游戏引擎。
Rich worlds, layered narratives, philosophical questions and distinctive systems—as well as NPC AI, rendering, upscaling and game engines.

所以,我的推荐算法大概比较难伺候 😂
My recommendation algorithm has unusually specific tastes. 😂

全部仓库 · Repositories ↗ · Hugging Face ↗ · X ↗ · Bilibili ↗


贴近硬件,让智能走向端侧。
Build close to the hardware. Bring intelligence to the edge.

Pinned Loading

  1. m3axpi_model m3axpi_model Public

    C++ 26 6

  2. DesktopPet DesktopPet Public

    Rust desktop pet with Mocari and Tauri

    Rust 1

  3. ax-AI_develop_guide ax-AI_develop_guide Public

    6 1

  4. ModuleLLM_Development_Guide ModuleLLM_Development_Guide Public

    C 5 3

  5. YOLO26.axera YOLO26.axera Public

    Python 5

  6. sherpa-onnx sherpa-onnx Public

    Forked from k2-fsa/sherpa-onnx

    Speech-to-text, text-to-speech, speaker diarization, and VAD using next-gen Kaldi with onnxruntime without Internet connection. Support embedded systems, Android, iOS, HarmonyOS, Raspberry Pi, RISC…

    C++ 4 1