#webllm — Public Fediverse posts

Phát triển AI 100% trên trình duyệt chơi Pokémon Red bằng Qwen 2.5 1.5B và mạng nơ-ron #AI #Pokémon #TríTuệNhânTạo #MạngNơRon #TrìnhDuyệt #Game #PokémonRed #Qwen #WebLLM

https://www.reddit.com/r/LocalLLaMA/comments/1ql6cz7/built_a_100_clientside_ai_that_plays_pokemon_red/

🚀 Ra mắt Focus Patrol: Extension Chrome giúp tăng tập trung, 100% xử lý AI cục bộ!

Sử dụng Gemini Nano hoặc WebLLM/Llama làm phương án dự phòng. Dữ liệu duyệt web của bạn sẽ được xóa sạch vào nửa đêm hàng ngày.

#FocusPatrol #ChromeExtension #AI #GeminiNano #WebLLM #Llama #Privacy #Productivity #CôngCụHỗTrợ #AIĐịaPhương #BảoMật #NăngSuất

https://www.reddit.com/r/SideProject/comments/1p1cbts/built_a_focusboosting_chrome_extension_using/

Fine-tuning mô hình trên groupchat: Qwen2.5 0.5B chạy trong trình duyệt
- Dược đào tạo từ 50,000 tin nhắn trong groupchat đại học
- Sử dụng Qwen3 4B, Qwen3 0.6B và Qwen2.5 0.5B để thu nhỏ mô hình
- Chạy trong trình duyệt với WebLLM
- Phủ nhận: có thể trò chuyện tại infinitegroupchat.com (cần WebGPU/iOS26)
- Hướng dẫn chuyển đổi mô hình sang định dạng MLC qua Colab

#AI #ModelFineTuning #Qwen #WebLLM #LocalLLaMA #MachineLearning
#TríTuệNhânTạo #ĐàoTạoMôHình #Qwen #WebLLM #LocalLLaMA

In-Browser Graph RAG with Kuzu-WASM and WebLLM — https://blog.kuzudb.com/post/kuzu-wasm-rag/
#HackerNews #In-Browser #Graph #RAG #Kuzu-WASM #WebLLM #Technology #Innovation

In-Browser Graph RAG with Kuzu-WASM and WebLLM — https://blog.kuzudb.com/post/kuzu-wasm-rag/
#HackerNews #In-Browser #Graph #RAG #Kuzu-WASM #WebLLM #Technology #Innovation

In-Browser Graph RAG with Kuzu-WASM and WebLLM — https://blog.kuzudb.com/post/kuzu-wasm-rag/
#HackerNews #In-Browser #Graph #RAG #Kuzu-WASM #WebLLM #Technology #Innovation

In-Browser Graph RAG with Kuzu-WASM and WebLLM — https://blog.kuzudb.com/post/kuzu-wasm-rag/
#HackerNews #In-Browser #Graph #RAG #Kuzu-WASM #WebLLM #Technology #Innovation

ICYMI: Local chatbot capabilities arrive for web applications, reducing data privacy concerns: WebLLM and new browser APIs enable offline-capable chatbots that keep sensitive data on users' devices while maintaining functionality. https://ppc.land/local-chatbot-capabilities-arrive-for-web-applications-reducing-data-privacy-concerns/ #Chatbot #DataPrivacy #WebApplications #WebLLM #BrowserAPIs

ICYMI: Local chatbot capabilities arrive for web applications, reducing data privacy concerns: WebLLM and new browser APIs enable offline-capable chatbots that keep sensitive data on users' devices while maintaining functionality. https://ppc.land/local-chatbot-capabilities-arrive-for-web-applications-reducing-data-privacy-concerns/ #Chatbot #DataPrivacy #WebApplications #WebLLM #BrowserAPIs

ICYMI: Local chatbot capabilities arrive for web applications, reducing data privacy concerns: WebLLM and new browser APIs enable offline-capable chatbots that keep sensitive data on users' devices while maintaining functionality. https://ppc.land/local-chatbot-capabilities-arrive-for-web-applications-reducing-data-privacy-concerns/ #Chatbot #DataPrivacy #WebApplications #WebLLM #OfflineFunctionality

ICYMI: Local chatbot capabilities arrive for web applications, reducing data privacy concerns: WebLLM and new browser APIs enable offline-capable chatbots that keep sensitive data on users' devices while maintaining functionality. https://ppc.land/local-chatbot-capabilities-arrive-for-web-applications-reducing-data-privacy-concerns/ #Chatbot #DataPrivacy #WebApplications #WebLLM #OfflineFunctionality

Local chatbot capabilities arrive for web applications, reducing data privacy concerns: WebLLM and new browser APIs enable offline-capable chatbots that keep sensitive data on users' devices while maintaining functionality. https://ppc.land/local-chatbot-capabilities-arrive-for-web-applications-reducing-data-privacy-concerns/ #Chatbots #DataPrivacy #WebApplications #WebLLM #BrowserAPIs

Local chatbot capabilities arrive for web applications, reducing data privacy concerns: WebLLM and new browser APIs enable offline-capable chatbots that keep sensitive data on users' devices while maintaining functionality. https://ppc.land/local-chatbot-capabilities-arrive-for-web-applications-reducing-data-privacy-concerns/ #Chatbots #DataPrivacy #WebApplications #WebLLM #BrowserAPIs

Local chatbot capabilities arrive for web applications, reducing data privacy concerns: WebLLM and new browser APIs enable offline-capable chatbots that keep sensitive data on users' devices while maintaining functionality. https://ppc.land/local-chatbot-capabilities-arrive-for-web-applications-reducing-data-privacy-concerns/ #Chatbots #DataPrivacy #WebApplications #WebLLM #BrowserAPIs

Local chatbot capabilities arrive for web applications, reducing data privacy concerns: WebLLM and new browser APIs enable offline-capable chatbots that keep sensitive data on users' devices while maintaining functionality. https://ppc.land/local-chatbot-capabilities-arrive-for-web-applications-reducing-data-privacy-concerns/ #Chatbots #DataPrivacy #WebApplications #WebLLM #BrowserAPIs

Local chatbot capabilities arrive for web applications, reducing data privacy concerns: WebLLM and new browser APIs enable offline-capable chatbots that keep sensitive data on users' devices while maintaining functionality. https://ppc.land/local-chatbot-capabilities-arrive-for-web-applications-reducing-data-privacy-concerns/ #Chatbots #DataPrivacy #WebApplications #WebLLM #BrowserAPIs

Check out my three-part article series about the benefits of on-device large language models and learn how to add AI capabilities to your web apps: https://web.dev/articles/ai-llms-benefits #GenAI #WebLLM #PromptAPI

Check out my three-part article series about the benefits of on-device large language models and learn how to add AI capabilities to your web apps: https://web.dev/articles/ai-llms-benefits #GenAI #WebLLM #PromptAPI

Check out my three-part article series about the benefits of on-device large language models and learn how to add AI capabilities to your web apps: https://web.dev/articles/ai-llms-benefits #GenAI #WebLLM #PromptAPI

The future of publishing is empathic: tech that senses your emotional state, adapts tone, pacing, & design for safe, meaningful engagement—in vulnerability. Gentle, human-centered UI for a hyperconnected world. #EmpathicDesign #DigitalHumanism #CompassionateComputing #WebLLM

The future of publishing is empathic: tech that senses your emotional state, adapts tone, pacing, & design for safe, meaningful engagement—in vulnerability. Gentle, human-centered UI for a hyperconnected world. #EmpathicDesign #DigitalHumanism #CompassionateComputing #WebLLM

#開源分享 一款高性能的瀏覽器內LLM推理引擎：WebLLM
直接在Web瀏覽器中進行語言模型推理，無需伺服器，通過WebGPU進行加速，相容OpenAI API，可在本地使用開源模型

特點：
1、支持JSON模式結構化生成
2、支持多種模型，包括Llama、Phi、Gemma、Mistral、Qwen等
3、支持自訂模型集成、即插即用
4、支持串流媒體，即時輸出生成，適合聊天機器人和虛擬助手等互動式應用

專案地址： github.com/mlc-ai/web-llm

#webllm #llm

@smach It would be better to use an in-browser #LLM using #webLLM, #Transformers.js, #MediaPipe or window.ai rather than to send your data to #OpenAI.

@smach It would be better to use an in-browser #LLM using #webLLM, #Transformers.js, #MediaPipe or window.ai rather than to send your data to #OpenAI.

@smach It would be better to use an in-browser #LLM using #webLLM, #Transformers.js, #MediaPipe or window.ai rather than to send your data to #OpenAI.

@smach It would be better to use an in-browser #LLM using #webLLM, #Transformers.js, #MediaPipe or window.ai rather than to send your data to #OpenAI.

@smach It would be better to use an in-browser #LLM using #webLLM, #Transformers.js, #MediaPipe or window.ai rather than to send your data to #OpenAI.

@tomayac Great article! Two questions:
1. How does #MediaPipe for #web compare with #ONNX Runtime Web and #WebLLM?
2. When will it be possible to run Phi-3.5 (not Phi-2) on MediaPipe? https://github.com/google-ai-edge/mediapipe/issues/5357

@tomayac Great article! Two questions:
1. How does #MediaPipe for #web compare with #ONNX Runtime Web and #WebLLM?
2. When will it be possible to run Phi-3.5 (not Phi-2) on MediaPipe? https://github.com/google-ai-edge/mediapipe/issues/5357

@tomayac Great article! Two questions:
1. How does #MediaPipe for #web compare with #ONNX Runtime Web and #WebLLM?
2. When will it be possible to run Phi-3.5 (not Phi-2) on MediaPipe? https://github.com/google-ai-edge/mediapipe/issues/5357

@tomayac Great article! Two questions:
1. How does #MediaPipe for #web compare with #ONNX Runtime Web and #WebLLM?
2. When will it be possible to run Phi-3.5 (not Phi-2) on MediaPipe? https://github.com/google-ai-edge/mediapipe/issues/5357

@tomayac Great article! Two questions:
1. How does #MediaPipe for #web compare with #ONNX Runtime Web and #WebLLM?
2. When will it be possible to run Phi-3.5 (not Phi-2) on MediaPipe? https://github.com/google-ai-edge/mediapipe/issues/5357

As always, I had a great time at DWX! Here are my slides from my talk about local, offline-capable #GenerativeAI models via #WebLLM, #PromptAPI, #WebGPU, and #WebNN: https://www.thinktecture.com/contributions/generative-ai-power-im-web-progressive-web-apps-smarter-machen/ #pwa #dwx2024

As always, I had a great time at DWX! Here are my slides from my talk about local, offline-capable #GenerativeAI models via #WebLLM, #PromptAPI, #WebGPU, and #WebNN: https://www.thinktecture.com/contributions/generative-ai-power-im-web-progressive-web-apps-smarter-machen/ #pwa #dwx2024

As always, I had a great time at DWX! Here are my slides from my talk about local, offline-capable #GenerativeAI models via #WebLLM, #PromptAPI, #WebGPU, and #WebNN: https://www.thinktecture.com/contributions/generative-ai-power-im-web-progressive-web-apps-smarter-machen/ #pwa #dwx2024

#Blogged: Solving bundler issues when adding #WebLLM to your app https://labs.thinktecture.com/solving-bundler-issues-when-adding-webllm-to-your-app/ #ai #genai #generativeai #webapps

#Blogged: Solving bundler issues when adding #WebLLM to your app https://labs.thinktecture.com/solving-bundler-issues-when-adding-webllm-to-your-app/ #ai #genai #generativeai #webapps

#Blogged: Solving bundler issues when adding #WebLLM to your app https://labs.thinktecture.com/solving-bundler-issues-when-adding-webllm-to-your-app/ #ai #genai #generativeai #webapps

#Blogged: How to deal with “cannot find function” errors after updating #WebLLM. https://labs.thinktecture.com/solving-cannot-find-function-errors-after-updating-webllm/ #ai #genai #generativeai #webapps

#Blogged: How to deal with “cannot find function” errors after updating #WebLLM. https://labs.thinktecture.com/solving-cannot-find-function-errors-after-updating-webllm/ #ai #genai #generativeai #webapps

#Blogged: How to deal with “cannot find function” errors after updating #WebLLM. https://labs.thinktecture.com/solving-cannot-find-function-errors-after-updating-webllm/ #ai #genai #generativeai #webapps

are there any applications out there using #WebLLM?

are there any applications out there using #WebLLM?

are there any applications out there using #WebLLM?

are there any applications out there using #WebLLM?

are there any applications out there using #WebLLM?

#Blogged: Adding AI functions to Angular apps: local and offline-capable (German content) https://www.thinktecture.com/ai/ai-funktionen-zu-angular-apps-hinzufuegen-lokal-und-offlinefaehig/ #angular #webllm #generativeai

#Blogged: Adding AI functions to Angular apps: local and offline-capable (German content) https://www.thinktecture.com/ai/ai-funktionen-zu-angular-apps-hinzufuegen-lokal-und-offlinefaehig/ #angular #webllm #generativeai

@m_drabe @Giliell ja, denselben Punkt hat auch Sam Altman kürzlich gemacht. Gleichzeitig sind die Anstrengungen im Hardware- und Softwarebereich enorm, KI effizienter zu machen. Die Firmen wollen den computational load nicht auf dem Server haben, daher wird es wahrscheinlich Hybridlösungen geben mit TPU chips in allen Gerätekategorien und Modellen, die halb lokal und halb auf Servern laufen. #ParzivAI ist ganze Größenordnungen kleiner als zB #chatgpt und kann lokal laufen (vielleicht auch bald im Webbrowser (via #webLLM), kombiniert mit einem #RAG erfüllt er so seine Funktion.

@m_drabe @Giliell ja, denselben Punkt hat auch Sam Altman kürzlich gemacht. Gleichzeitig sind die Anstrengungen im Hardware- und Softwarebereich enorm, KI effizienter zu machen. Die Firmen wollen den computational load nicht auf dem Server haben, daher wird es wahrscheinlich Hybridlösungen geben mit TPU chips in allen Gerätekategorien und Modellen, die halb lokal und halb auf Servern laufen. #ParzivAI ist ganze Größenordnungen kleiner als zB #chatgpt und kann lokal laufen (vielleicht auch bald im Webbrowser (via #webLLM), kombiniert mit einem #RAG erfüllt er so seine Funktion.